How to make video ads that sell #115

Hello and a warm welcome to the new season, as you could almost call this. After the summer break, we're on the third episode, the second with you, Fredrik. In the first one, I covered a bit of what's happened since the summer, a big update. In today's episode, we'll talk a bit more about Meta and Meta advertising. We'll dive into an ad account, so stay with us. We have a new segment where we analyse Facebook Ad Library, TikTok or websites. Now we can start with Alba Health. We're in Facebook Ad Library. For those of you who can't see, we'll try to explain as clearly as possible, but you can hear…

No, we won't. Go and watch the video instead! If you're interested in the segment where we give ads a thumbs-up or thumbs-down, what's working right now and concrete tips, you need to see what Fredrik is sharing on the screen. Otherwise, fast-forward about ten minutes and you'll have the next segment.

Okay, let's start with the first ad. For background, you've chosen a company. We might do one company per episode or every other episode, and a few ads that we comment on. Why do we think it works or doesn't? What's good and bad in the ad, and what might I have done differently?

That'll be good. You can hear we're improvising at a high level here. All right, Alba Health does some kind of health thing, as you can tell from the name.

We shouldn't reveal too much. For the people listening, it might be the first time they see this, as if a Facebook ad turns up in their feed and they've never heard of the company.

Okay. We've probably mentioned it before too, but I don't think so. It's called ‘The nutritionist reveals her best tips for eczema’. ‘My name is Johanna…’ That's what we see on the screen. Start then, go ahead.

‘My name is [unclear in the Swedish source: Johanna Holt]. I'm a nutritionist and health coach. Eczema affects up to one in five children, and many parents struggle with flare-ups, creams and sleepless nights. Eczema isn't just a skin problem. The root cause may be in the gut. When the gut microbiome is out of balance, it can trigger inflammation that may manifest as eczema, asthma and digestive problems. Have you tried creams, dietary changes and supplements without real change? Get to the bottom of the problem with Alba's test.’

Then it starts again. Okay. We can actually divide the ad into different parts. First, we have the hook, the first two to, say, five seconds. Does the opening make you stop scrolling? Both yes and no, I'd say. What's good is that they have a clear callout.

You mean ‘The nutritionist reveals’? It's a bit clickbait, her best tips for eczema. Now we've seen the whole ad, and there aren't any best tips for eczema. I haven't seen a single good tip for eczema, but it's good clickbait. The hook wants to get you into the ad.

Good or bad? I'd give it perhaps six out of ten. What do you think? Yes, the question is… You see a person in the feed. I like that part itself, but the question is whether you should visually show eczema right there. Do you see what I'm thinking? I understand exactly. I absolutely think you should have a stronger visual hook. Just text and a person feels relatively ad-like. Not hugely, because it isn't high-production. But you can clearly see: I'm a nutritionist and health coach, and now I want to sell something. That's the selling in itself. That's why I give the hook six out of ten. But ‘The nutritionist reveals her best tips for eczema’ is…

…a good callout that gets the right people interested in the video. If you have eczema, you'll watch. Exactly, if you have eczema, you might stop…

…you don't have to stop. No, but perhaps, and that's a good callout. The negative might be that if you don't have eczema, you scroll on. Many of Alba Health's customers probably don't have eczema…

…but the ad goes into depth for those who do. What I'm thinking is that it increases conversion among those who actually have eczema.

A hundred times over. In that sense, it's a good callout, which is why it's six out of ten and not lower. The hook isn't good in itself, and it starts quite slowly. The sound isn't entirely optimal. Not much happens on the screen, as you say. You don't see anything. Not much is happening in terms of…

Can we play the beginning, just the first five seconds? ‘My name is Johanna… I'm a nutritionist and health coach. Eczema affects up to one in five children, and many parents…’

If she starts with ‘My name is Johanna’ and ‘I'm a health coach’, that's quite boring. A hundred times out of a hundred, I'd have started with something else, such as ‘Eczema affects…’ It's easy to start with an introduction in storytelling or in front of the camera. It's natural to introduce yourself when you enter a room and don't know the person on the other side of the screen. ‘Hello, I'm…’ Then you want to give the background or say what you do. The credibility is being a health coach or nutritionist. You start with credibility and raise the bar: why should you listen to me?

I understand why you open that way, but you have to go straight into the problem. ‘Here are my best tips for eczema. Hello, my name is Johanna…’ Or ‘Would you like a good tip on how to cure or manage eczema, or do X, Y and Z?’ If it were acne, you'd do the same thing: ‘Hello, my name is Johanna and I'm a nutritionist…’

‘…I've studied…’ Or give your little background: I've worked with… Because that's where credibility comes in. But the first hook needs to relate to the callout on the screen.

Exactly, it should relate and match. Not starting with ‘Hello, my name is Johanna and I'm a nutritionist. Now you should listen to what I'm going to say about eczema.’

The hook is the most important thing in a creative, a video ad. Absolutely the most important, by far. If you have a bad hook, the ad never performs. With a good hook, it performs better than it otherwise would. The rest of the video and the USPs at least get a chance, because with a bad hook, you've already scrolled on.

If we take the next step, she says ‘Eczema affects…’ Yes. ‘Eczema affects up to one in five children, and many parents struggle with flare-ups, creams and…’

Here the ad gets much better. If it was six out of ten, it goes up to eight. For one thing, she brings up statistics. From experience and comparable ads, we know that increases hold rate. Both credibility for the video and for her as the person communicating the message. Here it's educational; she wants to educate us about eczema. They've done the script well. They bring it up and show visually that it's one in five. With little sounds, both picture and sound, you get the feeling that you want to keep watching. Otherwise, if you just saw her in front of a…

…camera that never changed, it would be quite a boring video. Now there are little sound effects and B-roll, meaning you visually show what you're talking about. Like the child running on the screen, because it affects a lot of children and young people. Then they move into ‘Have you also tried creams…?’ to create that recognition.

The script itself is bloody good. Here we go from six out of ten for the hook to eight out of ten for hold rate. In other words: you bring people into the video. Of those who came in, how many choose to watch for at least fifteen seconds?

If we take the next segment, she has ‘Eczema isn't just a skin problem’. Yes. Then someone may have tried all these different things. Exactly.

Oh, so it isn't just the skin that's the problem. ‘Eczema isn't just a skin problem. The root cause may be in the gut. When the gut microbiome is out of balance, it can trigger inflammation that may manifest…’

This is a good continuation. Now you've got me into the video, told me why I should watch and why you're credible. Statistics back it up: one in five can be affected. Then you move on to showing what the root cause could be.

It's relatable for many parents struggling with eczema who have surely tried loads of creams. Not us versus them, but an alternative to everything else being sold to you on the market. Why what you've already tried might not work: perhaps the cause isn't on the surface but hidden in the gut, hard to detect. If you haven't tested the microbiome yourself, it could be hiding there. Now we know through research that a lot shows up in the microbiome, right? Then they show it visually, so the visual effect reinforces what she's talking about.

Still a very good video and a good script. Still quite a subdued voice and monotonous. If the voiceover or delivery had been a bit better, it could have gone up to nine out of ten. Right now, I'm keeping it somewhere around eight.

What would she have needed to do? Change the dynamics of her voice? Yes, the speed too. Dynamics, speed and more happening when she's in front of the camera than just sitting and reciting a script.

Should she show something visually, hold something up…? No, she doesn't have to. But here you've filmed in front of a wall.

Film her in the living room, in a different setting, while she's moving, or holding the phone and moving around. A different angle, then?

Yes, then things happen. It doesn't have to be a different angle in front of the same background, I think. Instead, she moves around, a change of setting…

…and not just once, but several times. Then something happens. We can also see that the script has been cut several times and the sound level goes up and down. It doesn't feel coherent; it doesn't flow well. The editing might be seven out of ten. You'd want better flow, so it feels like one take. Now you can hear the breaks, which isn't ideal. People may perceive it as less credible and drop off or not click through.

Another tip: sometimes you hear popping sounds in videos. It's a defect, or whatever it's called in the editing software. If you add a small, gentle fade, a microsecond, the cuts aren't as obvious. They get a little faded and aren't perceived in the same way.

Yes, I had no idea about that. You're hearing podcast producer Fredrik speaking. Tips from the coach. Let's continue. This is the ending now. They talk about it being able to trigger inflammation, and then you see…

…pictures of people with eczema. ‘…that may manifest as eczema, asthma and digestive problems. Have you tried creams, dietary changes and supplements without real change? Get to the bottom of the problem with Alba's test.’ Here, at the end, you can clearly hear how poorly edited it is when it switches from Johanna as a health coach to Alba, the actual USP or call to action. The CTA doesn't fit with the video. It feels like two different takes, which it is, put together badly. Then the whole experience becomes second-rate at the end. The start was a six, up to eight, then down to six again because the CTA doesn't fit with the video…

…and could be much more specific. It should have been done in the video. Here I have a tip for people recording scripts, since I've filmed so much: when you write, plan where the B-roll and A-roll will be. That comes from film reels in the old days, I think, when you had physical reels.

One film reel was pointed at you, and B-roll, I guess, was the second reel that you pasted over it. That's my guess. Okay.

That sounds right. What I'd like to add is that instead of trying to remember everything, you can use a physical teleprompter in front of the camera. Then you don't have to remember the script if you want flow.

There are bloody good apps for that now. Exactly. But to some extent, it shows, depending on the camera angle. It can show that you're not looking into the camera. Like when I'm not looking you in the eyes but at, I don't know, your belly button.

I get it. There are apps to fix the eyes too. But otherwise, you can plan that at these points we'll show things visually. Then the person doesn't need to remember the text. When I film people, I let them hold the script in front of them and read it out.

That way, you get more flow. The most important thing is not to lean down, because then, if you have a lav mic here, you're speaking…

…straight into it. Then you can't lean over and read, because the voice changes. Even now, as I do it, I'm speaking into the microphone. Your voice changes both because of the position of your throat and because you're speaking straight into the microphone.

I get it, I agree. Bloody hell, so many good tips from the coach. But overall, the ad: six out of ten for the hook, eight for hold rate, seven as you get into it, and then six. The overall rating with the visual copy is six and a half to seven out of ten. It's very niche.

Both the product itself is aimed at parents, and specifically parents whose children have those problems. You're filtering out a large part of the population. That doesn't have to be negative; it can be positive if it's done well. Somewhere between six and a half and seven is what I'd give it. Are you reading the whole copy if you click it, or does it end there? Yes. The visual hook was one thing…

…but here it says ‘How many hours have you spent googling your child's eczema?’ Not bad, because parents with that problem probably recognise themselves. Instead of reinforcing what we see and hear in the video, they have different copy, which is positive overall. You can do the text copy in two ways.

Either you have what's actually been said, then USPs and a call to action following AIDA. Or, like Alba Health, copy that's probably been tested on other ads, worked well there and now works here. ‘If you're still looking for answers, it may be time to stop searching the internet and instead look at data from your child's own body. With Alba's gut test, you get a detailed analysis of your child's…’ The only thing really missing is a USP. I'd have liked the one-in-five statistic and a clear URL, a CTA: click here. If you're not watching the video…

…and it comes up in your feed and you read the text. Of course, there's ‘Order now’ and the ‘Order now’ purchase button. But while you're reading, you're quite far from the button. I'd guess the target audience is parents from 35–40 to sixty years old. Many see it in their Facebook feed and like to click directly on the CTA in the ad text. A URL is measurable, effective and increases the click-through ratio quite a lot. There's one missing here, and overall, almost all the ads we see in the library lack a URL in the text. There's just the purchase button, often further down and positioned differently depending on where in the feed you see the ad.

You get lots of tips in one here. The copy is good but could be better. It says: ‘With Alba Health's gut test, you get a detailed analysis of your child's gut flora, personal recommendations and support from a certified nutritionist.’

You think the link should go straight after that? Yes. Click here to read more, or click here to order the test for delivery to your home.

Then you have a URL to the test. Let's take the next one. It isn't a video but a carousel with different meal pictures and lots of text over food. I'll read: ‘Did you know your gut flora is as unique as a fingerprint? What makes one person's gut rejoice doesn't have to be what makes yours feel its best. With a gut flora test from Alba Health, you don't have to guess. You find out what's actually in your gut, and you also get a concrete plan to support it. Curious? For more information, go to…’ If we click there, we see Alba Health.

A tag there. I was happy at ‘go to’ and thought they were going to link to the site. From an ad, you basically want to link directly to the product page 99 times out of a hundred, for the easiest route from ad to purchase.

You don't want to lead people somewhere else in 99 out of a hundred cases. Occasionally you do, perhaps to an external landing page such as a newspaper article, with a longer route. Usually, you want to do that with products where the purchasing decision takes longer. Not Alba, but think sofas, furniture or Twistboxes, where the purchasing decision is longer or much longer than simply seeing an ad, recognising your problem and wanting to order the solution.

Back to the ad. It's a collaboration with Food Pharmacy, probably an influencer account with tips, information or education about food as medicine. Food Pharmacy speaks for itself…

Yes, they've written books and other things. I don't know that particular account, but in the ad we see educational material. ‘Did you know your gut flora is as unique as a fingerprint?’ No evidence, but okay. ‘What makes one person's gut rejoice doesn't have to make yours feel its best.’ No evidence there either. When there are claims like that, I'd want support, according to this or that, or anything at all. ‘Your friend eats porridge and feels energetic all morning. You eat the same porridge and are hungry before ten. How is that even possible?’ I don't know. No idea.

How is that even possible? Most likely it's everything else you eat around it. There's more to it. It's a hook in itself. You have oats and a lot of text above them. Then you're supposed to scroll on: ‘Researchers asked the same question. In one of the world's largest dietary studies, 800 people ate the same meals. The result…’

‘The body responded surprisingly differently to the same food.’ No, Sherlock. ‘Even identical twins reacted differently. Same genes, same food, different results. A clue lies in the gut. Identical twins share less than half of their gut bacteria.’ Of course. That's interesting in relation to tests and why I recommend various tests. I started feeling a bit worse and thought: what the fuck have I changed lately? I'd started adding vinegar, sauerkraut and things like that…

…which my AI knows. I discussed how I was getting more tired, feeling run-down and worse. It has all my lab reports and gut tests, everything, and said…

‘Have you possibly started having sauerkraut or vinegar lately?’ I hadn't said that. I was like: yes, why do you ask? From the lab tests, it knew I lacked the bacteria to break them down. I was eating something that's healthy for many people…

…but it just kept building up more and more, and I felt worse because I couldn't break down the substances. I'm sensitive to histamine.

An example of the ad's exact target audience. Fredrik, you'd have recognised yourself. ‘Healthy food is still healthy, but its effect on the body can differ depending on the bacteria already living in the gut. Something works fantastically for one person and does almost nothing for another. Same food, [unclear in the Swedish source: olika bak].’ An educational carousel with similar pictures…

…from July. I don't really know what to say. Not a bad ad, a good complement in the library, you could add. A carousel with good pictures, information and education is great. Not directly selling, but indirectly.

More informative than sales-focused. Ads like that tend to live longer, because people can relate, like them and send them to friends. Maybe we talked about having the same breakfast, but you were tired in the afternoon and I felt good. How come? Then you see it in an ad. You tend to see what you're talking about…

…for another reason, and when we see it, I want to pass it on because we've just talked about it. Now I've taken a bit more. The nice thing about Facebook Ad Library, if you want to learn about ads, is that you can see the date they've been running from. In this case, 9 July 2026.

Which platforms, which genders? Here the age is 25 to 65-plus, and EU ad delivery shows a reach of 147,449 unique people.

Exactly, the breakdown by age and gender too. Often, it says unknown in many cases, because perhaps people don't want to disclose it. But you have to…

Yes, you have to. But for some reason, not everything gets tracked. I think it's within the EU. In the US… I think the EU has some transparency…

…policy. Here it's mostly Sweden; the ad has run almost only here in Swedish. Not surprising, I think the company is Swedish and sells mostly here.

Shall we look at one from June? I'm scrolling a bit; is there another ad that makes you want to click it?

We can click any of them. If we go back to this ad, I basically think it's bloody good to have a carousel in the library.

It serves a purpose there. I tend to have a carousel in almost every other ad set, because variety is good. Video and still images are all well and good, but carousels have their place in the library. The fact that it's educational and informative, not just directly selling, is positive. A lot of it is good. I'd give it perhaps seven out of ten because the copy is missing a little, and the imagery is repetitive, with lots of substantial text per image. A better hook at the beginning could catch more people and make them even more likely to keep scrolling. Seven out of ten from me is still high. I'm not saying it's bad…

…as I said, it has its role in the library and should be seen as part of the whole alongside other ads. I haven't seen them all, but from the quick scroll, it felt like a positive contribution. Interesting that the site doesn't automatically take you… It asked about Swedish but doesn't switch. It really feels… On the website, that is.

On the website. We won't get into that; we'll look at the ads. Let's take one last one. I also want to add something about the carousel: the first image with the zoom-in…

I didn't understand until I looked at the others that the first was a very zoomed-in picture of oats. Neither did I. Next: ‘Tired of guessing and googling? Get an initial insight into your child's gut health with our quick quiz. Free and based on research.’ Then ‘Free gut health report, learn more’. More ‘learn more’ than ‘order now’.

Press play. Shall I enlarge it? ‘If your child is constipated and laxatives aren't helping, here's why. Parents are often told it's nothing to worry about, that they should just try laxatives or give more fibre. That can help in the short term. But for many children, the problems return again and again. That's because recurring constipation isn't always caused by diet. Often, it may be linked to an imbalance in the gut microbiome. With Alba's free gut analysis, which takes just two minutes, you answer ten questions about your child's symptoms and get personal insights into what may be behind the constipation.’

‘If your child is constipated and laxatives aren't helping.’ I think it's much better than the first ad. A better hook. The voiceover feels personal: is your child constipated and laxatives aren't helping, here's why. Extremely niche again. The parents who recognise themselves will probably watch to the end because they think they're learning something new.

Again, educational and informative, not sales-focused. Very good. They skip ‘Hello, my name is Johanna, I'm a nutritionist or health guru’. In that way, it's better. The voiceover is faster and more to the point. Much better. The picture fits the text: is your child constipated? You see a child on the toilet…

…which fits with the text and voice. A very good hook, easily eight out of ten. Even a child sitting and pulling at [unclear in the Swedish source: kulen].

And it moves quickly to the next sequence. A lot happens, which is good in ads, because people's attention spans are short. I usually say something has to happen on the screen within two seconds. Otherwise, you want to scroll on; attention doesn't last longer than two seconds.

Here you have both the visuals and the sound in the voiceover making you want to stay for the next sequence. Then the next, and the next, and the next. Things happen nonstop.

Play it and we can look at hold rate. Let's go from the beginning: ‘If your child is constipated and laxatives aren't helping, here's why. Parents are often told it's nothing to worry about, that they should just try laxatives or give more fibre. That can help in the short term. But for many children, the problems return again and again. That's because recurring constipation isn't always caused by diet. Often, it may be linked to an imbalance in the gut microbiome.’ Pause. Again, a fast pace, but variation and a dramatic pause, a micro-pause where it's needed. Emphasising words like ‘imbalance’ makes you want to see more of…

…the ad. It feels like whoever did the voiceover is a professional specifically in ads. Obviously a skilled creator they've hired or have on the team. The text is a little duplicated. ‘Microbiome’ in English at the bottom and the Swedish version too. ‘Come on’ down there; it doesn't feel like it fits with…

That must be a mistake in the subtitles down there. They don't fit with what's being shown. Exactly.

The B-roll doesn't belong with the ad. I agree, the editing is second-rate. It should have been removed here. But it is what it is and, in a rough assessment, very good. However, the CTA comes in the middle: ‘Do our free…’ while you're talking about something else. Wait until the end, otherwise you're not following AIDA: attention, interest, desire, action. You're cutting it off halfway through, with a subtitle that's wrong for what's being communicated on the screen.

It feels more like they missed that something else was written underneath. Or they had access to certain material they needed to show, bought footage they hadn't produced themselves, or made it with AI that inserted the text.

It says ‘Oh, come on [unclear in the Swedish source: barka]’. Completely irrelevant to the communication. The ping sounds are good, but come one or a few milliseconds too early relative to what she's saying. The editor could have done better. Careless mistakes rather than anything else. But if one creator made the entire ad without an agency, that's a good creator. You might pay one to two thousand for something like that.

If that's what you've paid, I think it's bloody good and gives value for money. I'd order more if the creator had done the script and visual material. If there's an agency behind it—I don't know who's running Alba Health's ads—I'd think there were too many careless mistakes…

…and that they should pull themselves together and do a bit better. The careless mistakes should have been spotted. But overall, a very good ad, better than the first one with the nutritionist. Definitely a role in the library. The B-roll is good at the end too: the app is shown, the gut flora test you're supposed to do is shown, and what it is on the screen. It seems easy to carry out, and then you get answers. It feels simple, and I can imagine it's easy to drive clicks from the ad. It's a free test being sold here, not the whole Alba Health test kit, just ‘Do our test…’

‘…and you'll get help to see whether your child also has constipation.’ Now I click on ‘read more’ and land on a quiz in English, I think. A little odd with English, but ‘Find out what's really going on in your child's gut’. Good landing page, simple. ‘Ten quick questions.’ You know straight away it's only ten quick questions, then a personal mini gut report. Good, perfect, good enough. Bloody good for ads and an initial assessment that becomes a lead. I assume an email address is required to get the report, and then you can communicate with the same customers by email…

…which is better than ads all the time. I think the quiz is fun, getting the email address and being able to follow up and educate over time. It's been running since 26 June, and we're recording on 24 August. A few weeks, about two months. Perfect. I don't know the ad spend…

We can see roughly how many they've reached. It's targeted at 18–65, basically everyone, all genders, with a reach of 74,000. Not a huge amount.

Not a huge amount; it could also be wrong. I've seen it off by a factor of ten sometimes. It may have reached 75,000 unique people ten times each in two months.

Then perhaps you have 750,000… But it's a unique thing: a Swedish ad for parents whose children have constipation.

No gender or age restriction, so technically everyone in Sweden can be reached. But it only reaches 74,756 in two months. If we scroll to who it's reached most, it's women aged 25–34…

…followed by women aged 35–44, definitely mothers. Women over 65, only a hundred people. People between 18 and 24, also only a hundred, so…

Here we see, Sweden, men aged 65, 39 people. Almost nothing. Despite no age or gender limits, it's probably being targeted at mothers in midlife…

…with young children. Meta's AI understands that. Again, you don't need an age or gender restriction to find the right people. We advocate that at Porobic Group: broad, general ads and not restricting AI too much. If you do, the ad performs worse.

But how do you think about the upper funnel and those closer to purchase, then? Say the first phase is just supposed to get people to stop. Someone doesn't know this solution to the problem exists.

Upper, mid and bottom funnel, the whole funnel structure still exists, but nowadays it's in the ads' creative. All the ads we just saw are…

…upper/mid funnel. You're not selling directly but indirectly, and making people aware of the problem. I like that. Bottom funnel, purchase ads with campaigns and products, belongs in the library too…

…absolutely. But a little of the ad spend should go there, maybe five to fifteen–twenty per cent in total. Like DPA: you looked at a T-shirt on our site and get it in your face…

…as a reminder that it exists or is now reduced in price. That sort absolutely has a place. But the majority should go towards information, education, storytelling, good voiceovers and B-roll with a good CTA for a test or information. Share free knowledge in ads, like here. It's appreciated. Overall, they're doing a good job.

Room for improvement? Definitely. In tough competition, with what's probably a low budget, there are always improvements to make. But there's no question they're doing a good job.

How do you usually divide it up? You can set it up so that those who've never known about the problem, or who know they have a problem but can't put their finger on it, should first just stop in the feed. Then you can arrange for those who've watched 25, 75 or a hundred per cent of a video to get the next video. That, in turn, can address questions they have or something similar.

In the past, we could build almost an entire funnel: if the ad is doing bloody well with one, two or three million views and you keep running it, you might want to give everyone who's seen it a follow-up video. That might only get 50,000–100,000 views, but that's enough, because it converts much higher. The first was informative and educational, the next more direct and straight selling. That's how we could do it before. Now, running both in parallel is enough, and Meta runs the first one first, because the AI understands in the background that it…

…should run first and this one second. I read that Nvidia's CEO, I'm going to butcher the name, Jensen Huang, said Meta is applying AI better than any company in the world right now. I believe that. When I see the strength of the ad engine and how good it's become at segmenting and personalising, there's no need for me as a creative person or media buyer behind the screen to divide it up: this one first, and after it, you should see this one. We saw that even without age or gender restrictions, ads are shown to relevant people.

Even those 39 over 65 could be relevant, because they have grandchildren with eczema or gut problems and can pass it on. A little more upper funnel, one step further from the problem. Or they themselves have children with problems; there are people over 65 who do. I think the AI is good in the background. As a media buyer, you shouldn't sit there tinkering too much…

How many creative divisions would you have? The first ad catches their attention at all, the second is educational and informative. Then perhaps one that addresses all the objections, whatever that's called in Swedish?

I understand what you mean, but no, I wouldn't have it like that. It depends on what you're working with. Mockberg doesn't need that kind of library, but aspirational advertising: I see a piece of jewellery on a person who looks like me, or someone I'd like to be more like. It would suit me, and I want to order it. Nudient, Ninepine or Stronger also need more aspirational content. Såpa or Alba Health need more information and education: here's the problem, here's the solution. Problem-to-solution brands. You have a problem, here's part of the solution or the whole solution. Then you also need a broad spectrum of different…

…creatives for different problems. People are in different segments. ‘Problem unaware’ people don't even know the problem exists and need information ads. Many parents who have children with problems don't know the gut flora could be a problem or that there's a solution. They're completely unaware. Then there are those who know the gut flora can have problems…

…but that's a small proportion. They're choosing between your report, perhaps Werlabs or going to the doctor for a thorough examination…

…or like us, testing almost everything imaginable. Different audiences need creatives that appeal to them in different ways. Right? I usually say variety in the library is good, so ads aren't copy-paste or too similar to each other. The voiceover script we saw could have run in ten variations back in the day. Now one is enough, good enough. You don't need ten visual creatives or hooks where you only change one small part of the text. In the past, a best-performing ad could run in ten to fifteen variations to squeeze everything out of it. Now you can…

…reuse parts, such as B-roll, but the creative itself needs to be completely different. You want the hook, the first two to three seconds, to be as unique as possible.

Then we're done with Facebook advertising. That turned into a substantial masterclass in Facebook ads, really. I hope you got practical tips. If you have questions about the three ads, or work at or with Alba Health, contact me on LinkedIn and we can discuss it in more depth. I like the new segment you came up with, Fredrik, going into libraries and looking. I hope those of you listening watched too. We're on YouTube…

…and publish everything in video format. Spotify has video now too. One fun thing coming is… ChatGPT once said, it's not new and perhaps many people already know…

…that they'd never have ads. Now they're launching, or have launched, advertising in Sweden. For larger agencies right now, because you have to apply. At this stage, not just anyone can sign up. Worth starting to investigate. You can already register an account, but I don't think you're allowed to advertise if you're just a [unclear in the Swedish source: startare]; you have to apply to get started.

What's exciting is that early movers can stand out, because so few are allowed to advertise, and even fewer get started and know how.

Honestly, I think it's a cry for help. OpenAI is bleeding money. They have a load of free users and notice B2B users going to Anthropic's Claude or other apps. They're saddled with the mass of people who want OpenAI for free. Few of that huge number subscribe, and even then it's twenty dollars a month.

It's not a huge revenue stream for the company. They're bleeding money and thinking about how to bring more in ahead of an IPO, at least to show proof of concept: now we have ads. These come after a generated answer and should be clearly marked sponsored, perhaps not even relevant to the answer. I don't know how good that will be for end users, with a limited selection of advertisers and ads only now and then. How do you pay, CPM or…? No idea.

It'll be interesting. Now we're showing visually what it looks like, and there's more to read on their website. The ad comes after an answer. Do you see it there?

I see. With so much data they've collected about the free users, it should be close to purchase. I do quite a bit of shopping through ChatGPT myself.

Absolutely, but if the answer is unrelated? Here it says ‘Ads do not influence the answers you get from ChatGPT. Your chat stays private.’ So the ad is unrelated to the answer. If you ask for the three best supplements you should take—I work with personalised supplements—and perhaps have the gut flora problem Alba shows…

…and need something, it recommends one thing and shows an ad for something else. No, it doesn't say that. Only that it should be clearly labelled as sponsored.

Yes, under ‘Visually separated’. Look here. No, under groceries, under the ad. Not there. In the ad. Zoom into the phone on the left.

Here it says ‘Ads do not influence the answers you get from ChatGPT’. Exactly, the answer is still personal. No, that means the actual answer you get isn't influenced by the ad.

Exactly. But the ad is influenced by what the answer is. I buy that. If you ask for the top ten vacuum cleaners, it lists ten…

…and then the eleventh comes up as an ad. Yes, exactly. Completely unrelated to these, because you asked for the top ten. The eleventh can be sponsored, and it can be a vacuum cleaner. But I'm thinking about the likelihood of you clicking sponsored content when it's so clearly marked and you can see it's an ad. I'd say it's very far from purchase, because you want what ChatGPT personally recommends, not what someone has paid to recommend to you. But maybe I'm wrong. I'm also thinking about how ChatGPT tends to change its answers depending on how you prompt. I often tend to ask…

Okay, which is best? It's common to google ‘What's the best’. Among the most common things. But lately, I've started asking what reviews they've received.

What reviews, and what are people complaining about? When I was going to change gyms, I wanted to know what people were happy with and what they complained about.

Then it could show whether it's the service… No, they'd renovated, and people didn't like how it looked afterwards. That doesn't have to mean the gym is bad. Those who liked the old one didn't like the new one. When I was looking for a camera tripod, the first recommendation said it was the best, but the reviews said it breaks really fucking quickly…

…so it also depends on the prompt and recommendation. We don't know; nobody knows. Highly speculative. We know OpenAI is bleeding money and has introduced this. The beta test had a number of hand-picked advertisers with million-dollar budgets, around H&M in size, and it went on for over a year.

They've hinted at ads and had it on the roadmap for 2026. Now it's being rolled out, including in Sweden. Just a handful, by application, so still on a large-company basis.

It's not inconceivable that Västerbottenssåpa will soon test ChatGPT ads. Will we do it? Of course; we like testing and being early with new formats.

I'm not saying we shouldn't do it. I don't know what the practical effect will be. ChatGPT is bottom funnel, close to purchase when you've done a lot of research, but I don't think the ad will make me click on Såpa and buy.

It also says here that the signals include context and intent in the conversation, the ad's landing page, title and text, and what the advertiser has shared. Anything else would be strange. We have to assume the ad should be as relevant to the prompt as possible. But it won't be like that at the beginning, with few advertisers and lots of people prompting.

Then you see one ad in a hundred prompts if it's hyper-specific. At the beginning, they want to serve ads to many people, and then they won't be that relevant. With a bigger library and more advertisers, they become more hyper-personalised.

There's no question they have user data, otherwise the AI wouldn't be so good. But we'll see; that remains to be seen. I don't think it'll be an effective channel for e-commerce businesses, especially zero-to-a-hundred brands. Hard to see, but I could be wrong. I have been before. Time will tell.

I say I have no idea. We have to try it first, really try it. But an interesting aside and a good continuation of the conversation about Meta advertising. Going back to that, a few major updates came over the summer while we've been away. Broadly speaking, Meta is moving towards an automated business for e-commerce businesses and advertisers, away from media buying. Agencies and media buyers get less say. Exactly what we've been flagging for years on the podcast, what we've talked about before. The biggest change is that more manual placement exclusions are being removed.

What we saw in Alba Health's library: no exclusion by either gender or age, but the ad is shown to the right people in the right target audience.

Exactly, they're removing it now in August, so you can't exclude. It's become more usable, in other words. The big change. What do you think? I think they're trying to make it easier because they can now. It's no fun sitting there excluding and including. If it can happen automatically, that's convenient.

Some media buyers have probably liked manually excluding the Instagram feed when the signal says it's performing badly. That's not needed now, because AI sees if it isn't working after a hundred impressions. Then we show less there and run Stories because that works, or Reels if that works better for this creative.

With that said, Meta has become smarter and performs better when it chooses for itself. Exclusions have really been a bad idea for years, working against Meta's AI. But aren't they removing Messenger Stories too? If a format really doesn't work…

…it also disappears as an option. I actually didn't know it was going away. But good to phase it out if it no longer exists. One thing I'm worried about, where Meta has made a change, is Advantage+ campaigns. Most of the library is probably Advantage+ today. The campaigns and ads there now mean Meta can change text in the ad image itself.

Not the copy above it, which is quite easy to change, but in the image itself. There, you might have an overlay or hook, as in Alba Health's carousel.

Now AI can change the text. Frightening, I think, but it says two things. One: keep a very close eye on Advantage+. With fact-based or educational ads, you need to understand how the change could be misinterpreted or go wrong against brand guidelines. Two: the creative is raw input, and Meta determines the output. Are you okay with that as an advertiser, or does it conflict with guidelines? You're still responsible but give AI the freedom to change text.

It makes my stomach tighten. A good copywriter or advertiser writes good text, and the fact that the algorithm can now change it feels scary. Especially at the beginning, it'll be shaky. I don't know if I'd have enabled it. If possible, I'd have switched it off, or run less Advantage+ because of it.

AI usually gets good at what it's trained on. If they say the new version is really good at security vulnerabilities, it's no coincidence; it's been trained on that kind of data and become good at…

…that thing. Meta has an advantage there, with an enormous number of ads to train on. It could become fantastically skilled.

I just think that initially, before it's sharp enough, there's a big risk of something being pieced together that isn't desirable. I agree. The next update is that ad ranking uses more advanced sequence learning.

It looks at the sequences and analyses each video individually and alongside other ads in the same account. Then it applies which should be shown first, second, third, fourth, fifth and seventeenth to the same person…

…and analyses millions of ads from thousands or tens of thousands of advertisers: which ones suit Fredrik best? Also, the right moment. If you're on the underground, in the car or at the office, different types might be shown…

…depending on the context. It's become much better at understanding the content and showing the right ad to the right person. We see trends in the account. An ad gets very little spend, has been there for two days, then a bit more spend and a bit more. More and more visibility, because Meta's sequenced learning has judged that it might fit better at the bottom of the funnel in this ad set. Right now, another should be shown higher up and before this ad…

…and that's why you shouldn't tinker often. Exactly. A common mistake: you see no effect, lower the budget, pause and say ‘Fuck, it doesn't work.’ Leave them alone. For how long?

In proportion to the money spent and the product's price. Smart. I usually use 1.5 times CPA as a good guideline.

If the product costs, say, three thousand kronor—[unclear in the Swedish source: som allält kostar 23.000]—you should perhaps spend 1.5 times what it costs to acquire a customer. Suppose a customer costs a thousand kronor. Then perhaps you should spend fifteen hundred on the ad before you have reasonably good data to pause or remake it. Per creative, then?

Per creative, yes. A rule of thumb; everything can be refined. I also want to add that Messenger Stories isn't being removed from the account, only from the Messenger app.

Where are they going? I don't know; I don't use the feature. I just read that. Do you know what's craziest?

What? I've worked with Meta ads for absolutely ages, hands-on in accounts, run strategy teams and an agency that works with it. But I deleted the Facebook app many years ago, haven't used the Instagram app and barely use Messenger except for family. I'm no longer a consumer of Meta Platforms…

…not a big consumer, anyway. Crazy, right? I have something similar and only get one minute on Facebook a day.

Ugh. My timer. Blocked in the phone's basic settings? Yes, and I've built my own desktop software. Crazy. Fucking nice, that's where you want to get to. I think I'll do the same. It's harder on mobile, but on the computer…

…I've made exceptions for Ad Library, Facebook Business and things like that. I'm there to produce, not consume.

Smart. A good change in July is that we have an MCP for Meta ads. The agent can connect directly to Meta Ads, an expected launch. A lot of people have wanted an official MCP from Claude or different agents, so they can give instructions and get a media buyer that more or less does what the agency used to do.

We've tested it in a… We talked about it in the last episode. I think this is part of the future and that more and more advertisers will adopt it…

…and have an agent buying media to a greater extent than historically. A good update. A lot at Meta is about us feeding in raw material. AI will edit material, add overlays, A/B test on a large and small scale and change the raw material, while still sending the traffic to the right page. The question is how much longer it'll be what you consider the right landing page. Soon, perhaps you'll just feed in a URL…

…a bit of raw material, then Meta does the rest. That's probably their stated vision: creating creatives, even AI creatives that are credible enough and convert better than what you've made yourself. Better copy and traffic to different landing pages that are A/B tested. One is better, so more traffic goes there. Meta is moving very quickly in developing Meta Ads. But still think from the basics. I'm only assuming now, but Claude doesn't have access to the whole of Facebook's ad library, or…

Now it does through MCP. But not the whole library to train on the data? Right. No, you can give access to your account.

Exactly, the model becomes based on the account. You can teach Claude, but you can't get an agent fundamentally, say Opus five or…

No, you can't train it yourself on all the ads in the ad library. Meta has an enormous advantage there. Think of everyone advertising on Facebook and all the data. I think…

…quite quickly, that model can become very good. No, I think I agree. [unclear in the Swedish source: Jag tror blir du klå att göra det då.] No, you can't. However, I know there are lots of third-party apps now…

…that claim to have been trained to identify good and bad ads. It's not technically impossible. At the agency, we also have ads performing well right now in Sweden for different brands, problem-to-solution and visual. There are different kinds of ads and creatives. Training an agent is theoretically possible, but then it has to be kept updated and continuously fed with data. Some third-party apps can work bloody well…

…as inspiration for your own agent. We've tested a few; so far, I'm not hugely impressed. But I think a lot will come this autumn, with the apps getting much better, before Meta's agent becomes fully good, perhaps one or two years away. It's now being trained on all ads, but a large part of the library is bad…

Do you think it's trained on all ads? I definitely think it's trained on all ads. If I were them, I'd train it the way I prompt: this is how I don't want it, and this is how I want it. That helps AI generate better.

At the same time, you want to give it freedom to make its own decisions and a large amount of data to train on. They can see which ads perform and which don't.

I know. Depending on the country, gender and where you live, there can be big differences, even between northern and southern Sweden.

A large population similar to the US probably works well, because a lot of training happens on American data first, where Meta Platforms' revenues are highest. It's natural to start in the US and tailor it to that, while little Sweden falls behind.

We'll see. I don't know. A lot of AI is still a black box; even OpenAI barely knows how many answers are produced. How will Meta's AI… Once you give it room to operate, you can't control it, and there are hallucinations.

Interesting to follow. A long episode; fun with Alba Health and account analysis, visually too. Thank you so much, Fredrik.

A strong comeback after the summer. Thank you to everyone listening. If you've listened all the way through and think it's interesting, leave a review. All comments and reviews make us very happy. Share it with colleagues, friends and acquaintances who would benefit from the segment.

Thank you so much. See you next week, now on Thursday as usual. Bye!