NOLL TILL 100 — ENGLISH TRANSCRIPT
Episode 120: How to scale towards 100 million with one employee #120

English translation of the full Swedish conversation. Original timestamps preserved.
Original audio: https://www.youtube.com/watch?v=_pxWwkuZ9xk

[00:00:00] A very warm welcome to a new episode of Noll till 100. My name is Fedja Porobic, and with me as usual is Fredrik Ankarsköld. New week, fresh start. How are you?

[00:00:12] I'm good. Today we're jumping straight into this week's topic: scaling brands with as few employees as possible, some sort of lean scaling. I know you've mentioned Twistboxes having a target of one employee at a hundred million. That was the idea when we invested in the company last year. So far we haven't hired anyone, and we're getting by largely through building AI functions that support Micke, the CEO.

[00:00:41] One of the ideas has been to build AI bots for outreach to buyers around the world. The product is a folding roof box. You might want to get your hands on it because it costs between ten and twelve thousand kronor, depending on where in the world you buy it.

[00:01:05] At that price, you might want to see whether it works. It's an innovation, not an iteration. There are hardly any folding roof boxes that actually work. Is it weatherproof and safe? Those are questions you want to ask someone. We've had an outreach bot for over a year that contacts both small and large buyers.

[00:01:30] When it gets a bite, it books a meeting with Micke so he can present and demonstrate it on a video call. First it qualifies a lead by asking relevant questions: what kind of store do you run? Do you have an email address? Do you have any revenue?

[00:01:55] We also tailor the material accordingly. It sends out PDFs and instructional videos on how Twistboxes works so people are informed before the meeting. Then you get fifteen minutes with Micke, the founder of Twistboxes. And it's a bit like, 'Oh, the founder.' You don't know how small the company actually is.

[00:02:13] In Sweden we're fortunate to have information on every company, and we can easily look them up. But if you're from Serbia, for instance, where a retailer recently bought from us, it's much harder.

[00:02:24] The idea is Micke, and then of course we have Porobic Group as the backend engine, with a team that technically works on Twistboxes. In principle, any brand could have that. The goal is to scale resources up and down easily without having to hire.

[00:02:46] Otherwise you're tied into long contracts and have to find work for colleagues who might not have a full workload. In an agency setup you can scale up: ahead of Black Week we need lots of ads, and someone needs to handle this. Then we need to scale down when sales are low or during the off-season.

[00:03:07] Maybe not that many people buy a roof box in January. It makes it easy for Twistboxes to manage. The fascinating thing is that people have predicted AI can't communicate. I thought of the craziest story I've heard when you said it handles the sales process up to the meeting.

[00:03:34] Some people told me about someone who built this for dating sites. Dating sites? Yes, he has AI chat with a number of people at the same time. When it's time to meet, he gets pinged and receives a brief about the person before the date.

[00:03:59] He's chosen that as a way to find someone to be in a relationship with. Okay, that's where we are. I haven't dated in a very long time.

[00:04:14] Neither have I, not for many years. That was before Tinder and all that. But it's crazy: the chat has a back-and-forth conversation with these people until it's time to meet, and then he gets pinged.

[00:04:32] It's not actually that difficult. You can train AI on how you speak, the language and vocabulary you use, so it makes it very realistic. I can imagine it becoming completely impossible for the people chatting to tell it's an AI.

[00:04:52] Or at least in ninety-nine percent of cases. I also think a lot of people on those sites use AI to help write their messages.

[00:05:03] That makes it even harder to tell them apart. Exactly. I don't know what to think. Maybe it'll end up being just a bunch of bots talking to each other.

[00:05:13] I think so. Elon Musk said somewhere that we're building for the next billion people, where that billion is digital people or agents. Soon there will be thirty billion digital agents in the world. Every individual with access to a computer will have several running around the clock.

[00:05:41] That's not unreasonable. I already have several agents running around the clock for the companies. I don't have any yet, but I actually spoke to my wife over the weekend.

[00:05:53] We were thinking: shouldn't we have an agent that creates the week's meal plan? You're always discussing whether we could come up with something new and what we're going to eat tomorrow. You don't want to sit there on Sunday evening asking what we're eating on Monday. An agent can suggest ten recipes for next week and know what we eat and what we ought to be getting.

[00:06:20] It knows macro- and micronutrition and can put together a weekly menu that gives you the week's nutrition, or whatever you'd call it. It should also reduce waste because it can see the number of calories and how much each person needs. Okay, but for me waste isn't the main focus; it's creativity and composing menus that suit us.

[00:06:51] Of course you can do some research and write down thirty dishes to rotate, but that's not what I mean. How do we make this week's soup, fish or meat into something new almost every time, without compromising? I have an extremely long list of things I don't eat.

[00:07:13] It should be an ingredient, not an ingredient list. An ingredient list is often a warning list. I stick to pure ingredients: no sugar or added sugar and no added E-number additives. Lots of things like that. My son is a bit picky about food. My daughter eats most things, so that's easy.

[00:07:36] How hard can it be to create an agent like that? Five, maybe ten minutes, and it's done. It sits there with its scheduled recurring task. It can also see whether there's anything else I need to keep track of and do new things. Like when I get an email saying the Serbian teacher is cancelling tomorrow's lesson.

[00:08:00] Then it goes into the calendar and deletes the lesson. Remind [unclear in the source: Lessia] that there's no Serbian tomorrow. A simple thing that otherwise takes mental capacity gets automated. A family agent, perfect.

[00:08:16] Why not? Speaking of AI and agents, we're building several for Twistboxes. One is the outreach agent, which has actually existed since the summer of 2025. About a year and a half.

[00:08:30] Even before that. No, more than a year and a half... It was launched in spring 2025. Almost a year and a half. It has booked hundreds of meetings that have led to fifty to a hundred sales. We're now represented in around fifty to a hundred countries.

[00:08:49] I'm so out of touch with it myself. But there's Japan, South Korea, Australia, Serbia, Panama, Mexico, the US, several European countries, the Nordic countries and China. Major buyers like Decathlon have contacted us and bought for several stores, physical and online. For a store, surely one advantage is that the products take up less space on the warehouse shelf than an ordinary roof box?

[00:09:27] Exactly. The next thing is that there are big differences between the personas buying the roof box. Someone lives in a cramped city-centre home and wants the largest possible roof box that takes up the least possible space. That hasn't been possible before. Now it is, with Twistboxes. We have a dedicated landing page for that.

[00:09:53] We run ads aimed at city-centre people with limited space at home. Don't just think Stockholm, but Japan, with many millions of people living in very small spaces. It's one of the countries where people camp the most. Campsites are booked up until 2030 throughout the country, not just the popular ones. They're then resold at secondary auctions, sometimes for millions of kronor.

[00:10:21] Completely sick. So you build a landing page for Japan, focused on people who live in small spaces and want room for their camping trip. You have specific traffic with ads for that need. Conversion on landing pages like that is somewhere between fifty and seventy percent better than on a generic page for everyone and everything.

[00:10:51] Otherwise, even if the ad targets the right person, the customer lands on a site with pictures of a Swedish family in the countryside, snowboarders or downhill skiers. The content points in all sorts of directions and tries to appeal too broadly. With AI today, you can simply have a content library.

[00:11:17] You have a basic template and Shopify MCP, essentially a connector through both OpenAI and Claude. Then you can set up a landing page in minutes. Say two hours while AI works and you make a few adjustments. Then you can drive traffic to different pages. At Breakit Ecom Day a month or so ago, I met a guy who had built AI specifically for that.

[00:11:52] At first I thought I ought to outsource it to him and save myself my two hours with AI. But we didn't do a deal. They wanted a couple of thousand kronor for each landing page, and I thought I could automate it myself.

[00:12:12] So we did it ourselves. A hot tip: if you know that different personas shop with you, which nine out of ten or even more ecommerce retailers ought to have, it's crystal clear.

[00:12:30] Use different landing pages and drive traffic there. I know myself that if I land on a site that isn't tailored to my needs, the likelihood goes down. At the same time, like many others, I have this idea that it doesn't affect me as much.

[00:12:51] Statistically, it affects you more than you think. You can feel that you're not the kind of person who's influenced, and still be influenced. I also listened to someone in content production. They spend forty million a year solely on producing content, not on advertising.

[00:13:10] Content. Okay, what brand is that? I can't name it here. A billion-kronor brand. What he was looking at was that even in their case, generic videos perform much worse, ones intended to bring in more people. In a longer video for a broad audience, they might mention different segments. Then they have another editor who picks out those short segments from the longer video.

[00:13:41] Do you follow? A longer video might be about how to use a roof box and show skiing, snowboarding and other things, depending on what you do.

[00:13:55] Visually. Exactly. An editor cuts out each segment, and those perform much better than the generic video. Completely reasonable; my observation is the same. When you get specific... Meta and other push media are good at finding the audience today, so you can be specific in your callout.

[00:14:23] 'This is for those of you who live right in the city, are short of space but still want to get away at weekends.' Something very specific. Or: 'This is for all horse riders who live outside the big city and own a horse farm.' Meta finds the people it suits.

[00:14:44] In fact, the more generic you are, the harder you make it for Meta. Depending on the product, of course. Some work for everyone, like soap: everyone has to clean.

[00:14:54] Then it's hard to have a callout: 'This is for everyone who loves cleaning.' But it could be: 'This is for everyone who hates cleaning.' Stefan, I hope you're listening. Everyone who hates cleaning.

[00:15:07] 'This is for those of you who love cleaning.' [unclear in the source: Scroll right, less.] It's different with decking soap and things like that. The idea is that with AI you can build different personas and have a written description of what each one does for work.

[00:15:26] It's easy to sketch this out with AI. You feed in: 'This is my persona. Develop different types and do customer research on it.' What pain points and lifestyle does the person have? Which points should you address?

[00:15:46] What tone should I write in? Exactly, but also what tone you shouldn't write in. That has helped when my AI produces content: knowing what it shouldn't do. Different copy gives different... When we worked a lot with iterations on Meta, we wrote copy aimed at younger people and older people respectively.

[00:16:09] Those are two different ads even with the same content, because we write in different ways. It worked much better than generic content for everyone. The same goes for makeup: if you're targeting an eighteen-year-old compared with a seventy-year-old, both the ad and the landing page differ.

[00:16:34] Before AI, it might have been too expensive for many brands, especially small ones trying to scale with few employees. It's a theme at an event I'm moderating in a week or so: Qliro's Ecom Summit. [unclear in the source: i brokslalt att göra.]

[00:16:52] Is that in Gothenburg? Yes. The event is about a term I call glocal: 'Globalization with a local touch' is roughly the theme. Lots of cool ecommerce retailers will be speaking. Above all, it's with the help of technology that this is now possible.

[00:17:14] The advice we're giving, which we use at Twistboxes, is what I'd use for most brands. You work in a similar way at Mockberg? Yes, exactly like that. We have a very small team.

[00:17:31] How many are you? We were about seven women last year, and Daniel. Then of course we have a backend.

[00:17:41] What do you mean by backend here? Things behind the scenes that can scale up and down: 3PL, outsourced production, systems, or Porobic Group as an agency for Meta, performance marketing, growth strategies and calculations. Resources that can be increased or reduced with the season.

[00:18:06] It becomes cost-effective, and you don't tie yourself down. Seven women scaled the brand to revenue last year of over a hundred... I don't know, over 150 million, very profitably. By the way, I think the figures are public now. A nice increase on the previous year and, above all, stabilisation of the company.

[00:18:32] We haven't piled on costs. We made a lot of mistakes in the first quarter of 2025, as I've talked about on the podcast, but then corrected them. What were the four most important mistakes?

[00:18:45] There was so much. Capital tied up in inventory is one of the biggest. In the fourth quarter of 2024, we invested heavily in the wrong products. We bought things we thought were trendy and cool and tried to create our own trends instead of doing what we're absolutely best at. Then we went back and [unclear in the source: pulled damage back]. We took on a lot of PR costs and things like that, which we could afford because we were profitable.

[00:19:14] But you always have to keep costs under control. It's better to reinvest in marketing and have cash in the bank than to tie everything up in stock and expensive PR, or things that might deliver something but are hard to measure. One key factor all along has been how we keep the team...

[00:19:41] ...as small as possible and do as much as possible with modern technology. Here's a tip, a fun challenge: for one working day, you're not allowed to do anything yourself. You have to use AI. Challenge yourself on every part; you're not even allowed to press a button yourself.

[00:20:03] What a challenge. So how do you press the button with AI? You have to ask it. It's to discover what you can automate. Then there are things where pressing it yourself is faster.

[00:20:17] A new podcast segment: Fredrik's challenge, the challenge of the week. Use AI for an entire day for everything you need to do: reply to messages and emails, press buttons and browse.

[00:20:37] You're not even allowed to copy and paste between emails and AI. You have to ask AI to go in, write and send the email. I love the challenge. It's inefficient. You'll lose a lot of time and get annoyed, but you'll also learn a lot.

[00:20:50] Exactly, that's the point. With some things I do, I've discovered AI might as well do them. Absolutely right. This is the challenge of the week, a good segment. Back to Mockberg and the small team.

[00:21:09] Back to The Lean Team. Another example we tested very early at Mockberg, back in 2024, was when we were entering Finland. We realised it was quite hard for us to explain things in Finnish because no one on the team speaks Finnish. So we translated Elvira's talking-head videos, what we call educational material, into Finnish.

[00:21:37] For those listening: these days it's not just dubbed, but lip-synced too. It looks as though the lips are moving in Finnish.

[00:21:46] Today, using various methods, it's impossible to tell it's AI. I use ElevenLabs, which we were already using back then. They have good natural [unclear in the source: prompters] for Swedish speakers. Small local languages like Swedish, and Finnish even more so, tend to be hard to make sound good.

[00:22:12] We solved that in 2024, so nine out of ten customers or people selling the products didn't realise she doesn't speak Finnish. Elvira was worried people would start speaking Finnish to her when she went there, when she doesn't know a word except hello. It worked out. A strong example of something that was possible then and definitely is now.

[00:22:37] Continuing with automation and tips, we do a lot of B2B sales at Twistboxes. We've contacted hundreds and sold to fifty to a hundred retailers. Then we notice there's a standard set of the same text and images, regardless of where the product is sold.

[00:23:06] Whether it's Japan, Sweden, Italy or Australia, they get the same texts. There are different seasons and variations, and people may have different preferences, but we have the same text everywhere. Okay, what do we do about that? If we sell into Serbia, AI can first translate it into Serbian, which is obvious. But it can also adapt both the content and copy to the country.

[00:23:39] Images, video and text. It's an adaptation, not a direct translation. For example, it searches for which cars are most popular and produces content where our roof box fits those cars. It already does; we really have built a product that works with all cars.

[00:24:00] So do you generate an image with the roof box on the most popular car and send it in the retailer pitch? Not in the pitch, but in the sales deck. When it's going onto the site, the store gets a Dropbox or Google Drive link with specially produced content: unique images and texts just for them.

[00:24:24] Everything gets adapted, even with small changes. The technical specifications must of course be identical. But the description, body text and title are slightly different. Then Google and other engines don't see it as duplicate content. They don't compete with each other; they get their own SEO ranking to compete in the local market.

[00:24:53] Here's another tip when creating content. That was a bloody good tip, but another one concerns product pages. Many people write about a backpack, for instance: 'This is a large backpack and it's waterproof.' I've noticed that in my case, I want to know more. Is it waterproof because the material itself is, or has it been treated? For example, I want to avoid PFAS.

[00:25:22] Of course, details. Yes, these details are very important for AI to know. Let's take an example: you want to optimise things so the retailers appear as suggestions in AI agents when their Serbian customers write and ask. The customer might say, 'I'd like a roof box,' and have various criteria. The more information there is on width, depth, height, litres and all the other variables the person might have in their specification...

[00:26:02] ...also needs to be on the retailers' product pages. Absolutely right. It sounds obvious, but I bought light bulbs the other day. On several sites, I had to click the photo and zoom in on the box to see whether my criteria were listed there.

[00:26:24] Didn't you want strong blue light? I wanted less blue light and preferably flicker-free bulbs, so the flickering doesn't affect my nervous system while I work, and also for when I'm filming. It's printed on the boxes in the photographs.

[00:26:43] But not in the product text. That's what AI picks up. It needs what Google has been asking for for a long time: structured data. That's easy to produce with AI today. You need to keep doing it for every product on all your ecommerce sites.

[00:27:07] Back to what you said: the health tip of the week. I read a Canadian study last week. I have this problem that if I drink too much coffee, I go to the toilet a lot. Caffeine is a diuretic. In my previous life, two years ago, I didn't have that problem. I ate a different diet, probably more restaurant food and more salt.

[00:27:35] Salt retains fluid, which my new diet doesn't do in the same way. I minimise the amount of salt, so it retains less fluid and I go to the toilet more. The study primarily looked at people over fifty who have more difficulty controlling their bladder and often wake up at night to go to the toilet.

[00:27:59] I haven't had that problem yet, but in the morning it feels as if I'm going to burst and I have to run to the toilet. One thing is to have a sip more often throughout the day instead of a whole glass of water.

[00:28:16] You spread out your consumption. Otherwise the water just passes through the body, through the liver, and you need to pee. If you drink one or two sips at a time, five minutes apart, it becomes easier for the body to absorb the water. The health tip of the week: drink more water, but more slowly and spread out more.

[00:28:44] And earlier in the day. The body needs time to process the fluid. I need to get all my fluids in before five o'clock. After five, I'm allowed to drink no more than one and a half decilitres for the rest of the evening.

[00:29:03] Wow, that's not much. Don't you feel dehydrated? You feel dehydrated, but you're not really, because I've had what I need during the day. If I stick to that, I don't wake up at night to go to the toilet.

[00:29:19] Whatever floats your boat. If I've drunk too little during the day and feel thirsty at eight o'clock and pour water into myself, I'm going to wake up needing the toilet.

[00:29:32] Then you wake up. Back to the study: among those over fifty, they found that smaller amounts of water more often meant retaining sixty percent more fluid in the body.

[00:29:46] Which is the point: retaining the water and not going to the toilet all the time. It can also mean you drink less because you retain more. A no-brainer. Back to lean scaling.

[00:30:03] Before that: I read on ehandel.se last week that Baby... Was it Babyland that went bankrupt? I always mix up Babyshop, Babyland and the other multibrand baby stores. Babyland has gone bankrupt after several difficult years.

[00:30:24] Last year they reduced both their revenue, not by a huge number of millions, but down to around 153 million, and made an enormous operating loss. Double digits, maybe fifty million. A lot of money, despite cutbacks and renting out parts of the warehouse, including to Västerbottens Såpa.

[00:30:53] We're in their warehouse in Morgongåva and have slowly increased our space. How is Såpa affected? I don't know, but I don't think it will hit us. Some operator will take over. A major owner of the baby store is [unclear in the source: Lalima], which I think is owned by Apotea, or Apotea's ownership group.

[00:31:22] It owns a large share of Apotea, which of course is listed and doing brilliantly. They've increased revenue from around 6.6 to just over seven billion.

[00:31:35] It was Babyland that went bankrupt. Yes, Babyland, sorry. Apotea is doing well, but less so for subsidiaries or other holdings in the portfolio, where Babyland has now gone bankrupt. Will new owners continue running it as a multibrand store, or will it disappear?

[00:32:02] My guess is someone will take over the bankruptcy estate and try to run it for another round, more efficiently. The question is whether you can make warehousing and logistics more efficient. That's more of a challenge. We're talking about how you build a brand and can actually scale it significantly. Warehousing is one part that requires more. It's a physical operation, and that can become a challenge.

[00:32:29] A huge challenge: you need large spaces for large products. Fewer children are also being born in Sweden, and fewer people are moving here following changes to asylum rules. There are fewer children to target, and especially in a few years it'll be even worse.

[00:32:53] It's tough. Many municipalities have to close preschools. But there's always room for skilled operators who find a clear niche and position and understand their reason for existing.

[00:33:10] Here I think it's a race to the bottom. They've positioned themselves as affordable with fast deliveries. Shipping a big sled or pram is expensive. A sled only costs three hundred, four hundred or five hundred kronor.

[00:33:31] But its dimensions make it expensive to ship. If you're a distributor or retailer, it's expensive: the manufacturer makes money from you, but you might not have the margin at the next stage to bid on Google Ads, Google Shopping, Prisjakt or PriceRunner.

[00:33:55] The audience is price-sensitive. If someone else drops the price by ten kronor, you can lose the purchase even though you've paid for the click. It's difficult, but I think it's possible with a clear position and niche. Instead of trying to reach everyone, you can target a specific audience within the baby sector.

[00:34:24] I don't know whether they ought to shrink revenue from 150 to perhaps fifty million but make it very profitable. Again, a lean, mean scaling machine, where one person runs everything. Then you need a warehouse team to build on top of that. It can be done, but will the next owner think that way?

[00:34:48] Or will they try to do what they've already done, optimise away a few costs and make it even more... No, I don't believe in that.

[00:34:59] I'm fairly certain it's going to be very tough, unfortunately. But if brands like that were created today, perhaps it would have been easier to build the systems from scratch for scaling. I can't comment on logistics, which seems to be a big part of the operation.

[00:35:22] One mistake several companies have made is building a very large logistics centre of their own and thinking it'll scale to the moon. If it doesn't, you're stuck with big costs.

[00:35:41] It's a recurring pattern in the industry. The reason was that they wanted a broad range because they saw a direct link between product range, revenue and profitability. More products gave them more keywords to buy. A new pram, or someone searching for a new brand...

[00:36:09] Then we have a way into the store. But you're not really building the organic engine where we get traffic regardless. Where we give people a good customer experience, get them to recommend us to each other and make them want to buy a second time. Apotea has succeeded phenomenally at more than a broad range and good prices. You get deliveries very quickly and consistently, with the right product, and have a good experience of the whole process.

[00:36:44] I don't think Babyland has managed to repeat that. Deliveries are a bit slower, there are more mistakes, and it isn't always the right price for the right product. They don't have a pricing robot that matches prices well. Even with that, it would have been an uphill struggle, because margins are poor for this type of ecommerce product.

[00:37:13] The packages look completely different: large, small, multiple SKUs or just one. It takes a lot of warehouse workers' time and is hard to automate. Såpa is simple: the products look roughly alike, and the packages come in few sizes. Apotea is similar: most things are small and fit through a letterbox.

[00:37:42] There are larger products too, but that's not the whole niche. A large part of their logistics can be automated, which isn't possible at Babyland. We'll see how the next owner wants to run the brand. When we talk about scaling with few employees, you can also look at...

[00:38:08] Charlie Munger talks about doing the opposite. Sometimes it's hard to get ideas about how to do something. If you instead ask how to really fuck it up, it's usually easy. In a relationship, you can quickly think of things.

[00:38:32] I don't use snus, but if I sat on the sofa throwing snus on the carpet, left the dishes out and assumed my partner would do all the housework, it would very likely not be good for the relationship.

[00:38:49] I hope people have the self-awareness to understand that. But it works in business too. Depending on the business you want to grow or already run, you can ask how it would work if everything went to hell, if scaling became really expensive and complex.

[00:39:12] You can split it into different categories. One is product development, one is advertising, then you have logistics and the website too. Take your case: you have Shopify. Then it's very easy to create a landing page with AI connected to Shopify, one for each audience.

[00:39:38] Then you create content that drives people to each page. That's easy to scale. How do we make it difficult? Choose a unique, outdated system that costs a lot and requires developers for every landing page. So it can be done in two completely different ways.

[00:40:05] Another tip for lean scaling is to automate reporting and media buying. Tracking takes an incredible number of hours. In the more serious ecommerce companies I've been involved with, there's been an entire tracking team.

[00:40:30] Sometimes you outsource to India, where they're very skilled, and the Philippines. I don't know about the Philippines, but I've worked with India several times. One of Sweden's biggest contact lens brands had a team in Sweden and an even larger outsourced team. Often the data tracking is scattered everywhere.

[00:40:54] You need good gut instinct to see how a customer becomes a customer. But gut instinct is based on experience, and that's replicable and scalable. ChatGPT, for example, has gained its own pixel and catalogue and a simple way for end customers to search for and find your products using structured data.

[00:41:20] Now, for once, it's possible to do performance marketing somewhat on autopilot. You can create at least mediocre content, or better, and have the data tracking. Start with how people reach the site and learn what traffic it is. Then the information goes back to ChatGPT's pixel, and you try to make small bid adjustments in Meta using ChatGPT or Claude.

[00:41:56] There's a direct connector to Meta and the ability to set rules for what's good or bad: ROAS targets, or preferably sales targets. Say I bid a thousand kronor a day on a product that costs three thousand and sell six thousand a day, but wouldn't have sold any at all without the bid.

[00:42:23] Now this is getting a bit advanced. Then you see a direct relationship between media buying and sales and can instruct the bot or agent to scale the account. It can work around the clock, which a human can't. Everything is data, trackable, and happens in front of a computer.

[00:42:44] There's nothing physical you have to go to a warehouse to check before scaling the Meta account. Everything in front of a computer can actually be automated. My call here is to automate, because there probably isn't a human with data tracking as good as the agent can have.

[00:43:07] It's always connected and can fetch several parameters from your Shopify collection every hour and use them in the media account. Maybe you shouldn't adjust ads and bids up and down all the time. But in zero to a hundred, once a day, or in a larger company several times, through the agent.

[00:43:33] Does that mean the agency's reason to exist for performance marketing disappears? I'd say that right now agencies need to be the ones setting it up. Then the agency's time is spent on other things. You won't sit in front of Meta buying media manually. Soon we'll see that as dinosaurs, outdated. Remember when we had to upload an ad, wait a few seconds, copy text from someone else's document, go back, select an audience, duplicate the ad and create it in another ad set.

[00:44:11] That took time. I recently spoke to a woman who wanted to help us: she could create the basic package, we'd do all the work and she'd upload it. I thought uploading isn't the hardest part today.

[00:44:30] That's simple. What's complex is what happens in between: creating content and planning what comes next based on what performs and what doesn't. Then there are different parts of the funnel, top of funnel and further down. What kind of content should be created?

[00:44:55] With the tracking you're describing, we can also get that data instead of just guessing. Exactly. One of the best indicators is, believe it or not, spend. I see media buyers time and again switch off an ad with worse metrics but higher spend in the same ad set, which an AI agent wouldn't have done.

[00:45:21] The agent understands that high spend can mean Meta's AI sees something the media buyer is missing. You're looking at a higher CPM, a higher [unclear in the source: CTA] and worse ROAS. You switch off the ad while four others in the same ad set have lower spend and look as if they're performing better.

[00:45:49] But all four die. The ad you switched off was bringing people down Meta's funnel. Maybe it was the first one shown, without being credited with the sale. Without it, there's nothing attracting people to the others, perhaps still images driving cheap traffic.

[00:46:14] The other ad looks as if it converts well: a higher click-through rate, lower CPM and higher ROAS in the platform. You think: 'I'm a genius. If I switch off the expensive ad, I'll get twice the results.'

[00:46:31] Then performance falls flat. You're sitting there scratching your head, not understanding what happened. 'We need new creatives,' you say. You get new material to work with, and then it all starts again. One of the single most important KPIs in a Meta account is precisely spend. The ad with the highest spend and worst metrics isn't always a bad ad.

[00:46:54] When I've experimented with agents, they seem to understand that better than ninety-nine percent of the media buyers I speak to. It's the most basic thing a media buyer should know. I recently filmed with Olivier Sibony, who co-wrote Noise: A Flaw in Human Judgment with Daniel Kahneman and Cass Sunstein.

[00:47:25] I can recommend the book. Among other things, they've studied judges and how many months or years in prison they give. It can vary depending on how they've slept, what they've eaten and whether it's before or after lunch. The sentence can differ by an extreme number of years.

[00:47:49] What? What I like about agents here is that... There's a black box with them too; sometimes you don't know why they do something. But you can design criteria and adjust them according to how the agent performs.

[00:48:12] It's easier to train an agent than a human, apparently. Yes, it has shortcomings, but before or after lunch might not be the decisive factor.

[00:48:25] Exactly. That brings us to the ad of the week. We haven't looked at it yet, but I've spoken to a guy whose startup is making the zero-to-a-hundred journey. It's largely subscription-based, not an AG1 copy, but roughly AG1 for the Swedish market.

[00:48:58] We'll look at the ads and what made the brand find some sort of product-market fit. It was slow for a while, then things took off, largely thanks to the creatives. Before we wrap this up: do you have three tips for building a company with few employees?

[00:49:29] Absolutely. One: look at what the minimum viable person is. What does a human actually need to do that requires a human touch, hands, feet or physical presence? Physically getting from A to B, doing something in the warehouse, appearing live in front of a camera, building community that way or organising an event.

[00:50:03] You can't outsource that. By that I mean agent-source it; I don't have a good word. Research, copy and the site, on the other hand, can be built. I've built a whole suite of applications for Porobic Group that can do [unclear in the source: fork]. We can cover that in the next episode.

[00:50:27] Then we can break down what we've built in days and how it makes things easier for many employees. Tip one: find what's essential for the brand, what you want to do and what can be agent-sourced. I'd like to add: dare to question everything.

[00:50:55] When someone says it's a feeling: a feeling is just a label for something we don't understand how to break down into actions. Like thinking someone is pleasant, which I've mentioned before: 'He was a nice guy.'

[00:51:13] That might be because he looks you in the eye, remembers your name and asks how you are and about everyday life. Then we label him 'nice' and say it was a feeling. But the feeling is based on actions the person, or the agent here, took. Tip two?

[00:51:33] To scale well, you or your AI need a complete handle on the data. Tracking should be set up correctly, structured data should be present and the site should be conversion-friendly. The catalogue should sell, not just say things. The product should have a clear audience and solve a real problem in someone's everyday life.

[00:52:05] The company needs a clear vision, mission and story. We emphasise the story a lot. People follow other people, not companies. I'd like to add here that you shouldn't get too stuck in your own story: read the book StoryBrand.

[00:52:24] Donald Miller. Yes, 2.0 has come out. We have to link the books you recommend; I've forgotten the previous one after five minutes. Put them in the show notes so even I can click. So: data tracking, accurate information and basic functions like Google Tag Manager and Google Search Console. Someone should actually look at it, even if it's an agent.

[00:52:51] So you can act on the company's data. Number three: what I've often called the backend in this episode. You should do what's at your core. If that's creating the best product, spend your time on that.

[00:53:13] Then product-market fit becomes your moat and what drives lots of sales. If you're very good at content, spend time on that. Do what you're actually good at and put a lot of the rest in a backend. You don't need to hire loads of people; you can work with agencies, external consultants, freelancers and now agents too.

[00:53:39] Whichever route you take, you should have professionals around you who complement you. At peaks you can scale up, in troughs down. The working model needs to be flexible, instead of an agency billing forty thousand a month regardless of how things are going.

[00:54:00] Things are going really well or really badly, but somewhere you're always paying for time. If you buy other people's services, you exchange money for time. If you buy AI companies' services, you're instead exchanging money for tokens, which are also some form of time the AI spends on your brand. That applies whatever brand you run. As a founder, agency owner, brand owner or CEO, you need to understand that the other side only spends time as long as it's paid for that time, or receives tokens for the time.

[00:54:42] Then the natural next step is to look at what that time is actually spent on. Can the agency's work be optimised, should you switch agencies, or do you even need one? I don't know.

[00:54:57] Exactly, do you need an agency? A refresher on Porobic Group: we're an agency, but very much a backend for brands I've invested in. You don't need to contact me. I can certainly recommend other good agencies, going forward into 2027. I get a lot of inbound interest from brands wanting to work with me after the podcast or stage appearances.

[00:55:20] I don't have capacity to take on any more new projects. If I do, I want to be a major shareholder in the brand. Then you need to give up a large stake at a reasonable valuation, which I'll probably push down because I bring value. If I'm going to wake up thinking about your brand, I want a big reason to.

[00:55:40] If we're going to build together, I want to do it over several years. A bonus tip about focusing on what you know: I often see people lack a basic understanding of... If I want to recruit an agency, I'll call it recruitment here...

[00:56:05] Say I want to recruit an agency. Then I need enough understanding of what I want: what should they achieve? I might have interview questions and interview several different agencies. Then the questions have to be the same for every agency so the comparison is consistent. Then I also have to understand how to measure the results. It shouldn't just be: 'They make more money,' but...

[00:56:29] In a meeting, I need to be able to understand. When we've looked at content creators for one of our brands, they say they can help us. If you ask how they usually measure success, you sometimes notice they haven't got a clue.

[00:56:48] Not a clue. Unfortunately, many people aren't keeping up. It's easier today: GPT Astra can give you a good basis for comparing agencies.

[00:57:01] You can have Wispr Flow running during the call, then put the transcript into Astra, or use any transcription tool. Some model listens in the background so the call can be transcribed. AI can then help ground your assessment in the conversation so it isn't just driven by feelings. Now let's move to the ad of the week.

[00:57:25] Fredrik, I've been looking forward to this segment. We're going to look at Relivo, which sells a product similar to AG1. For those who don't know, you could almost call AG1 a multivitamin, although it isn't. It's a scoop of nutrients for stressed people who might not have time to get them through real food.

[00:57:56] The question is how easy it is to get enough nutrients when the soils are being depleted. It's quite easy if you really... I'm thinking about the average person.

[00:58:05] You have to eat very well. Selenium is hard to get enough of if you don't eat Brazil nuts. For some people magnesium is difficult, for others something else. For me, the problem is that it's a generic product for everyone.

[00:58:25] People can have different deficiencies. I'm often sceptical of solutions like that, but we're not going to discuss the solution itself, just the ad. I only wanted to explain what the company sells. Here's the first ad, which was literally at the top of the Ad Library.

[00:58:46] When did it start running? The first of September, and now it's mid-October. One tip when buying supplements is to look at the ingredients. You may want to buy the product, but many cost a few hundred a month.

[00:59:02] It can even come to several thousand. Some people think blood tests costing several thousand are expensive, but you quickly save that amount. You can put the ingredient list into AI and ask what blood tests you need to take. There's Werlabs and Testmottagningen.

[00:59:19] And many more. I use Svensk Provtagning. There are several options. You can test whether you actually need the supplement. For example, I got excess selenium from a multivitamin and had to stop taking certain things. Too much isn't healthy either.

[00:59:40] Back to the ad. The text says: 'Time to fall asleep is at a record low.' Time to fall asleep? 'After two months with Relivo, Helen shows how the right nutrition actually affects your health.'

[00:59:58] Time to fall asleep is at a record low, meaning you fall asleep faster. What a poor expression. It's not something you normally say in everyday life. I wouldn't tell my wife I had a three-minute sleep-onset time yesterday.

[01:00:15] Maybe people who use an Oura Ring, Whoop or something similar. They don't call it sleep-onset time there either; it's usually latency. Okay, let's press play. 'What can happen if you take Relivo every day for a year?' A fairly good hook.

[01:00:28] It's written as text on the video. [Ad clip:] 'My name is Heléne Wikmar and I've been taking Relivo every morning for a year now, and I've had such incredibly good results. I notice it especially in my sleep. I fall asleep really quickly, and I really sleep... It only took a few days before I noticed the difference on my watch that records sleep, when I saw how the time it took to fall asleep changed from half an hour to an hour, to now three to ten minutes.'

[01:01:07] [The ad clip continues:] 'It's incredible. My deep sleep also went from twenty to thirty minutes to now around one and a half to two hours. It makes a huge difference to how I feel in everyday life, where I work full-time, exercise a lot and have three children to look after. There's something special about mixing a drink in the morning. It becomes the starting signal for the day.'

[01:01:33] [Ad clip:] 'I can really recommend Relivo. My name is Heléne Wikmar and I've been taking Relivo...' Okay, what do we think? To start with, the hook is very good: what happens if you take Relivo every day for a year? It sells a product that you shouldn't just take for thirty days.

[01:01:56] You're not promising loads of changes after thirty days, like many similar brands. Then people often try an offer and drop out after the first month. Here you're almost committing people to a year, to see what happens if they do it for a year.

[01:02:17] Like challenges: ninety days of calisthenics or push-ups every day for ninety days. What change do you get? It's a good video hook. Then she starts: 'My name is Heléne. I've been taking Revo every morning,' it says. Some words are literally misspelt.

[01:02:41] It's probably AI that... But it's so bad for the brand and has to be corrected. You can check the subtitles.

[01:02:51] That was poor. Then I think it's dull to start by introducing your name. Unless you're obviously famous, with several hundred thousand followers or frequently appearing in the media and press.

[01:03:10] Then partnership ads can be profitable. If you're obviously famous, the video can start with your name. Or if you have a very strong credential: 'I'm a [unclear in the source: physician in physics],' or a chemist who has dealt with these kinds of cases for thirty years.

[01:03:31] Or: 'I've worked with clinical sleep patients for twenty-seven years.' Then you immediately establish authority. Clearly it's about the time it takes to fall asleep: 'I've worked with insomnia for twenty-seven years.' That gives the video credibility. Otherwise I think you should go straight to the problem. Do you think people care what her name is?

[01:03:56] No idea. No, I don't think so. On the other hand, the visual hook is very good: she's on a boat, has a lovely background and there's movement. A lot is happening visually. The story is told with a voiceover recorded afterwards. She's made lots of little clips, edited them into a longer and quite pleasant video, then read out a script. An agency probably wrote it in the background, or she's a UGC creator herself and has sold something in.

[01:04:28] It's a weaker script. Not bad in itself, but not particularly good. It has a clear story, which is positive, but it's very salesy. I don't immediately get the sense that she's had sleep problems, even though she says the watch recorded her sleep-onset time, Oura or whatever smart device it was.

[01:05:02] It recorded a poor sleep-onset time, but I don't see any evidence. A screenshot of the year before and the year with Relivo would have provided something. More proof.

[01:05:16] Yes, you have a watch and can take a screenshot and show it. She mentions taking half an hour to an hour to fall asleep.

[01:05:27] Which improved to three to ten minutes, and recorded deep sleep of thirty minutes? Let's look. [Ad clip:] 'Deep sleep went from twenty minutes.' Twenty minutes consistently over time, then you have a major problem.

[01:05:47] Twenty minutes is almost nothing, so low. Here's a tip for those who are interested... What I mean is that Relivo isn't going to solve that problem.

[01:05:59] You have to see a doctor. Unless it was a deficiency that Relivo specifically helped with. I find that hard to see. Twenty minutes of deep sleep is basically dying.

[01:06:10] Twenty minutes is so low that it's bad. A tip for people with sleep problems: I was told to stress less and go to bed on time. I used to have a lot of sleep problems. I had blood tests and corrected the chain because there were deficiencies.

[01:06:32] When I corrected the deficiencies, I slept better. Sleep can be affected by nutritional deficiencies, excesses, when and how you eat, and light. We should partner with your lab, considering how much we recommend blood tests. I wish I got paid for them.

[01:06:52] I had a blood test yesterday, so I know exactly. Or at least got free tests. [Ad clip:] 'One and a half to two hours.' That's very good. Two hours is really good.

[01:07:07] Two hours is an extremely good deep-sleep pattern. Everything else matters too, but deep sleep is a measure of sleep quality. REM sleep also tells you something. Often the first cycle when you fall asleep is deep sleep. Then you move up into REM and alternate during the night.

[01:07:30] When you first fall asleep, the body goes straight into deep sleep to rest. That's why twenty minutes is very short. When and what you eat is also a challenge. For example, it's healthy to get a certain amount of carbohydrates, because that affects the body's ability to produce substances to sleep through the night.

[01:07:59] But is saying 'drink Relivo' a health claim? Or are we following someone whose subjective opinion is that she's taken it for a year? Above all, the ad targets other women who have trouble sleeping, particularly in midlife.

[01:08:25] There can be many causes of sleep difficulties. Here we're following Heléne Wikmar on a lovely boat trip. She's sitting on the terrace, having something that looks tasty and enjoying the sunshine. That's the feeling you're selling around a problem many people probably have, falling asleep. Now I'm calling it sleep onset too, just because the ad did.

[01:08:55] But what do you think of the ad overall? I'm thinking about how I interpret it. The music is almost meditative, and it seems calm and lovely. The pace isn't particularly fast, which fits with sleep. At the same time, she talks about exercising a lot and having three children to look after.

[01:09:19] Now she's alone in the archipelago, by the looks of it. Would that be useful? I don't know, but it's interesting. I don't think so; there's a lot of contradiction. It would have been better with b-roll of her stressed in the morning, showing the everyday life the product is meant to help with.

[01:09:40] Now you're selling a cosy moment in the sun with a glass of Relivo. I actually don't think the ad is particularly good and find it hard to see it working. But it's been running since September, over six weeks, and has obviously received spend and reached people. It has a reach of fifty-three thousand.

[01:10:02] I think Relivo chose it for lack of anything else, but I don't think it's good. 'For lack of anything else,' that's a bit funny. I really don't think it's that good. At the same time, it's a startup. Definition of done: you can always do better. How long should you wait for better material before publishing?

[01:10:23] Maybe just get the job done. In advertising, it's often hard to know in advance what will work. Companies need to experiment. There are fundamentals you can know, but through lots of experiments you get data points and learn. Then we're almost...

[01:10:43] Same company, another ad. What do you think? By the way, the previous one targets ages eighteen to sixty-five. That's just broad audience targeting.

[01:10:54] I think that's absolutely right, nothing strange about it. We barely need to comment after listeners have heard it for so long. But some might think that because there's a woman in the ad, it should target women. The product is also aimed at men.

[01:11:12] Many men can relate to sleep difficulties. They may also have daughters or a wife. If you scroll down, Meta has probably still targeted the ad mainly at women.

[01:11:30] There are actually lots of men too. Are there? Nice. Sort by reach. Here: female, female, male, female. The great majority are still women in midlife, thirty-five to sixty-four, as we said about the ad's audience with sleep difficulties.

[01:11:58] It also reaches some older men with sleep problems. I think Meta has done a phenomenal targeting job here. It's good to raise even the obvious things, even if many people know them.

[01:12:11] It's obvious to us. The next video has the tagline: 'Curious about Relivo? Here's how it works,' with an arrow pointing down. Take a scoop, mix with water and drink every morning. Each scoop combines thirty-eight carefully selected ingredients: vitamins, minerals, marine collagen, pre- and probiotics, berries and plant extracts.

[01:12:48] All in one simple daily routine instead of lots of jars. One scoop and one glass every morning. Made in Sweden, nice. Tested by Eurofins, which is a lab for third-party testing. And tested for substances classified as doping agents through [unclear in the source: a collagen list]. Then athletes can consider the product too.

[01:13:23] Many athletes don't dare take multivitamins and other supplements, especially from smaller companies, because they're afraid of substances classified as doping agents. It's understandable that they do a test. It costs a bit, I think fifteen to twenty thousand kronor for a company.

[01:13:47] The question is whether they test every batch. Otherwise they're not allowed to write that. You can't test one and not the next; you need to test every batch. It might consist of five or ten thousand products from the manufacturer, so it quickly gets expensive.

[01:14:08] But you have to do it if it's stated in the advertising. Consumers can request it, which I do. If someone writes that they have a COA, certificate of analysis, about the origin, I'll request it.

[01:14:27] For my cocoa, I requested it and was told they didn't have it for that batch, only another one. That doesn't help. It could come from another forest or harvest with more heavy metals. I want to know what applies to my batch.

[01:14:43] They had to backtrack and change the site and things. But that was also a big reason I bought the product. We know something similar from when we've ordered products from China. There are unscrupulous manufacturers there who say: 'This is CE-marked.' They can write on the factory website that the product is CE-marked.

[01:15:08] Then perhaps they've tested one of a hundred products and obtained CE marking, but not the other ninety-nine. It isn't the factory's job to do the testing. CE is a European standard for electronics allowing you to put the mark on your product.

[01:15:24] Do you mean when ordering white label? Even then. As the importer, you're responsible and can get into trouble. The importer has to make sure the product has a proper CE test in China or Asia. Otherwise you have to do it in Europe.

[01:15:41] To be able to sell it with CE marking in Sweden or Europe. Otherwise you can get into trouble. Few people actually check and quality-review that. I'm deeply involved in those kinds of tests too. But let's press play.

[01:15:57] We could soon call ourselves the Testing Podcast. [Ad clip:] 'Let me show you here. This is what the jar looks like. Here's the powder, you can see inside.' [Text in the ad:] 'Relivo explained in one minute.' [Ad clip:] 'Vitamins, minerals, pre- and probiotics, marine collagen. Take a scoop.'

[01:16:15] [Ad clip:] 'Top up with water. I love mixing it with this in the morning. No sweeteners. What do you love about it? It feels very natural when you drink it. Let me show you here. This is what the jar looks like.' Okay, it says 'explained in one minute,' but it's about twenty seconds. He also says 'no sweeteners.'

[01:16:37] I find it hard to see it having no sweeteners. Products like this usually have them to taste good. Or you use natural products, berry extracts here. Berries can be sweet. Click buy; the call to action is 'Order now.'

[01:16:57] If you click there, you see: 'Seven jars every morning, now one scoop.' They're selling replacing several different... Then you reach a site saying '15+ supplements in 60 seconds.' The ad said thirty-eight different ones, remember?

[01:17:17] Thirty-eight nutrients. Okay, it says that here too. You replace fifteen supplements with thirty-eight substances. I'm guessing that's confusing for customers too. Click 'Order now,' which flashes annoyingly, and you come down here.

[01:17:35] The button scrolls down to the starter kit. Let's look at the contents. Is there an ingredient list? That's just... Yes, an ingredient list, but you can't click the image to enlarge it.

[01:17:51] You almost have to zoom in. At the bottom right is allergy information, and here are minerals. It looks like a reasonable composition of mostly good things. Vanilla flavouring, malic acid and some [unclear in the source: emulator] are the negatives. But it says no sweeteners or unnecessary fillers, which is nice. I don't know how it tastes, but the flavour probably comes from the substances in the product.

[01:18:26] Here we see the sugar content. Carbohydrates two grams, 0.4 per scoop, I'm guessing. Not much sugar at all. A good amount of protein and fibre per scoop, but mainly minerals and vitamins. Fibre isn't the focus, although there are pre- and probiotics. Probably a better-for-you product.

[01:18:57] Do ordinary people need all the vitamins they might get through food? Probably not. Can you overdose? Probably. But we're not going to point at the product; let's return to the ad. 'Bring your daily nutrition together in one simple routine.' The product page, site and product image are quite good.

[01:19:24] They could be tailored more to the advertising with better landing pages, but that's a lot of work. Back to the ad: a product demonstration. It belongs further down the funnel.

[01:19:40] It's more retargeting. You don't actually need to set that in Meta, which will understand by itself. A bit of ASMR: here's the jar and here's how you use the product. It's easy to mix with the IKEA mixer. The jar is stylish and the powder looks fresh.

[01:20:07] The scoop seems overfilled. The demonstration might be wrong, but I don't know anything about that. The ad is good enough for an ad account together with the other one. The other is more top of funnel, and this one bottom of funnel. They probably work together.

[01:20:27] As you said earlier: a common mistake. This one might have higher ROAS, which we don't know, because it's reaching someone again who has already seen something else.

[01:20:39] Exactly. Here it's more obvious that it's retargeting. If it converts better, you might switch off the ads leading here, a common media-buying mistake. If we click 'Transparency by location,' we see the same audience but lower reach, which makes sense.

[01:21:02] Fewer people. Sort by reach. Still twenty-eight thousand. You probably see similar audiences but more men, because a man appears in the ad. It's actually fairly even between women and men. Roughly the same ages, with thirty-five to forty-four high.

[01:21:28] Women forty-five to fifty-four are also high. The insomnia ad had more older people; this one more younger people. A man appears, so it's targeted a little more at men. Meta is doing a good job again.

[01:21:42] Could it be better? A hundred times. This could have been a better emotional ad. A product demonstration at the cabin, by the water and the morning sun, would have given a lovely feeling: I start my day with Relivo. Now it's a bit dull and shut away in a flat.

[01:22:05] Couldn't that make it relatable? Not in this case. It's more about association, taking the product in the morning and creating a lovely feeling. The other one wants to show a concrete problem the product solves and introduce the brand. The ad does a good job in context.

[01:22:31] Maybe it doesn't need to be the absolute best, because it's reaching people again who are already interested, but it still has to generate clicks and purchases. Rating: the first ad about six and a half out of ten.

[01:22:44] With Heléne [unclear in the source: Wiklund], where the script, composition and contradictions bring down the rating. Still okay and possible to run. This one might get seven out of ten. Even though it's worse in itself and more quickly produced, it does its job for its purpose further down the funnel.

[01:23:17] When I see '15 supplements in 60 seconds' on the website, I think they're targeting someone who already takes several supplements and wants to combine them into one scoop. The ad about measuring deep sleep and time to fall asleep also targets people trying to optimise their health.

[01:23:45] If you click buy, you probably end up on the same part of the site. It's the homepage with the text about fifteen supplements. As far as I know, it doesn't say much about sleep, only simplifying the morning routine. The landing page doesn't mention sleep very much. If it instead reinforced the ad's sleep message, you'd probably have somewhere between fifty and seventy percent better conversion from that particular audience and ad.

[01:24:15] I'm reflecting on how the product section would normally show the product name, based on search. Here, 'Bring your nutrition together in a simple morning routine' is the size of an H1.

[01:24:36] It is an H1. But then perhaps you become less... It's not very good structured data for an agent. No, perhaps you don't search for 'I want to bring my nutrition together.'

[01:24:52] It's a typical performance-driven channel, not a search-driven one. Could it be made better for agents? A hundred times. I'd change the title and move the text somewhere else, perhaps the image.

[01:25:08] In this case I'd choose a landing page more optimised for agents and search engines. A hundred percent. Or this is actually the website: homepage or landing page. What they've done well, I noticed when I was in another store this morning. They had so few products, yet you still had to click and click, first getting to a category and then clicking on to the product page.

[01:25:34] They only had three products. Then I'd have preferred Relivo's solution with products available to buy on the homepage instead of clicking through an entire segment.

[01:25:44] Relivo could create separate landing pages. One ad was about sleep, so it could have a more sleep-focused H1.

[01:25:57] Images focused on sleep, a clear problem, a clear audience and advertising. A hundred percent. It shouldn't be hard to arrange. I think the founder listens and might fix it, and can get back to us...

[01:26:12] ...with how much better a conversion rate you got for that particular audience or ad. Yes, both are nice. What I want to praise Relivo for is trying to get rid of all the additives. People probably don't understand how many supplements out there have loads of additives. Sometimes when I'm in the pharmacy, I wonder: 'How are they even allowed to sell this stuff?'

[01:26:40] I do too. Now, this isn't the new Health Podcast. Fredrik, many thanks for this week. If you've listened this far, an hour and a half into one of our longer episodes, I know you want to hear the next one. Subscribe so you don't miss it. It signals to Spotify and Apple that you like the podcast and gives recommendations for similar podcasts.

[01:27:10] It's good for you and us and gives us motivation to carry on every week. We sit down on the morning sofa on Mondays, talk and share everything we've learnt and know. That didn't come out very well; we'll have to cut it. Or keep a blooper so you get to hear what it sounds like behind the scenes.

[01:27:33] I'm joking. Thanks for this week, and many thanks, Fredrik. See you again next Thursday. Bye!