Grow 50% without tying up all your cash in inventory #122

A warm welcome to a new episode of Noll till 100. My name is Fedja Porobic, and with me as usual is Fredrik Ankarsköld. A new week, a fresh start. In this week's episode… That sounds so fucking boring to say. We're going to talk about Black Week. We're right in the thick of it and about to enter what, for many of us, is the best period of the year. How do we avoid screwing it up? We already did a whole Black Week episode at the beginning of October. I'd really recommend going back and listening to it. If you haven't heard it, you'll get loads of useful information. If you've already listened, it's easily worth doing again. Repetition is good.

This week we're going through five not-so-quick points about Black Week. Then we'll move on to the main topic: forecasting for 2027, now made possible by AI. It was possible last year too, but back then we could only do very basic things. Now we can connect real data and get really good analysis and forecasting without having to guess. For those of you who stick around to the end, we even have a KPI scorecard that we use at Porobic Group and may possibly share with you.

Possibly! [laughter] We'll see if you stick around to the end. Good, straight into five not-so-quick points about Black Week. The first, and perhaps most important, thing: do you know your break-even ROAS, Fredrik?

Right, what did you have in mind there? [laughter] We did a calculation in the first Black Week episode, at the beginning of October. Do you remember the hypothetical case about what's left when you increase the discount from 20 to 30 percent?

Exactly. It's incredibly important to have a firm grip on that at this point. Factor in discounts, returns, shipping costs, tax, payment fees, and other predictable costs. What does your bottom line look like if you sell X, Y, or Z? A best case, a worst case, and a normal scenario. You set up three scenarios.

Quickly burning through all your profitability is a risk, especially during Black Week, when everything accelerates. I'd really encourage you to go back and listen to the episode or start doing the calculations in your own P&L, so you're on top of costs all the way down to the bottom line. A lot of people screw this up. I've been involved in so many cases where people say they've broken every all-time record. Yes, you have. But two months later, or actually two days later, you run into cash flow problems because you haven't accounted for the discount eating up all your profit and then some. Sometimes you even make a loss.

Very often, unfortunately. Occasionally it might be justified because you're acquiring new customers. But all too often these days it isn't, because discount-code hunters are hard to retain. That's really point number two: what do you do once Black Week is over? You've run heavy discounts, poured on the emails and probably text messages, and worked toward a period when you've brought in loads of new customers.

Many of those new customers are discount-code hunters. They bought the product because it was on sale. They might have wanted to try it for a long time and now had the chance at a lower price. Do you have a plan for when Black Week is over? This is where you need to start planning your retention strategy. How do we get first-time customers, whom we've spent a lot of money acquiring, to shop again? Introduce them to the brand. Include storytelling in your emails. Remember that they came in through discount ads, so they may be more fleeting than customers who came in during the year and often bought at full price.

I'm thinking of Groupon. Do you remember them? Yes, I remember Groupon. They're still around in the US. Exactly, but they shut down in Sweden. And even now they're struggling with a major restructuring…

…in the US. The company is still losing money. A lot of it is about bringing people in cheaply. Part of the pitch, at least, was that if you have a restaurant and bring guests in cheaply, you'll get customers who come back.

That's probably true, except people move on to the next restaurant offering something cheap. You saw that especially in service businesses. Massage therapists, hairdressers, tire retailers, and tire shops realized it and started adjusting their offers a little. Eventually the offers were so poor that you couldn't bring in new customers that way in a country as small as Sweden.

That's a hundred percent true. So, again: do you have a plan for how you'll work with the customers you've just acquired? What should that plan look like? One tip was to introduce storytelling once the Black Week frenzy is over.

Another could be showing people how to use the products. Include more educational and inspiring content in your emails and invite them into a community. You have a lot of room to communicate with the customer. You can send physical mail or include something in the first Black Week box the customer receives at home.

What could that be? A QR code that helps people get to know the brand on a deeper level. But how? The risk is that a business owner just throws in a QR code that leads to something uninteresting. What do…

…you define as interesting? What should the QR code lead to for people to actually scan it and engage? You have to think of it as needing to win the customer over again. You won the customer with a good offer, or a better offer than your competitors'. The person might have been curious and seen an ad before, but doesn't know the brand in depth. How do we bring the customer into the loop so they get to know the brand and use the products? A customer who uses the products is much more likely to recommend them, especially if they're good.

And I really hope we're selling things we think are good. Otherwise, I don't know why you're listening to the podcast. [laughter] I really don't want you here if you're listening in order to scam people…

…and sell something that scams people. Exactly. I don't want you in this podcast, so click away if you're scamming people. That's a good anti-hook. We should use that on social media. Fredrik, if you're someone who wants to scam people and grow your e-commerce company: don't click on Noll till 100.

Are you a scammer? Yeah, my God. If you're a scammer, keep scrolling. Next: what do you do if your ads stop performing? This is a really interesting discussion topic that I come back to on repeat. I know it'll come up on the Tuesday or Wednesday of Black Week.

Which key metrics do you look at to decide whether an ad is working during Black Week? Good question. The first thing I see is that sales take off as soon as you launch your Black Week offer. Whether that's Thursday, Saturday, or Sunday the week before, or Monday during Black Week…

…you get a sales spike straight away because you email all the customers who've been sitting there waiting. No-brainer. On day two, lots of people will still buy. On days three and four, you see a decline, because people shopped during the first two days.

That's not so strange, but we e-commerce people tend to increase ad spend on those days to compensate for the drop. We think we want to maintain that initial peak and keep the momentum going: we have a budget, we've done so much work, and we want to put our foot on the gas. So the budget increases to 100,000, maybe 200,000 kronor a day. If you normally spend 10,000, 20,000, or 30,000, you've suddenly tripled or quadrupled your normal daily budget.

This is really dangerous, especially if you don't have a grip on your bottom line. I saw a meme recently: ‘We've increased sales by 50 percent.’ Yes, but we've increased the ad budget by 400! [laughter]

It could also be successful. We don't know their numbers, after all. But the example shows that looking at one number at a time isn't always a good thing. Yeah, it's fascinating, this numerical sleight of hand and how people manipulate numbers in different contexts.

What we usually do at Porobic Group, for the brands we also have an ownership stake in, is have backups for our backups for our backups. What do I mean? We know that still images usually perform well during Black Week, and it's easy to produce new ones as backups. Your offer might not be as competitive as you thought. Twenty percent off the entire range might not work. ‘Buy three, pay for two’ throughout Black Week could be much more competitive.

It can give the customer a bigger saving while you make more on the bottom line. You need to do the calculation yourself, compared with simply offering 20 percent off your entire range. You can also achieve a higher average order value.

You can also encourage people to qualify for free shipping on the site. If the customer gets a discount, they still need to buy more to reach the free-shipping threshold, precisely because they've received a discount.

If the offer is poor, people won't buy. That's where you need a backup. If we haven't run the same offer year after year, if we're new, or if we have a new offer, we need to test it. Sometimes we stress-test the offer on 1,000, 2,000, or 3,000 people on the list during a flash sale, Halloween sale, or spontaneous autumn sale. It runs for 24 or 48 hours in early or mid-October. Then we have a rough idea of what to expect during Black Week.

How much of that applies during Black Week? You have the sales funnel with people who are already keen to buy. For them, you can have images that simply say: discount, come and buy. Then you have the top of the funnel, where people barely know the product exists or have only just started learning about the problem.

Say a company has a revenue target of 100 million. How many ads do they need at each stage? As an example, maybe ‘come and buy’ ads make up 5 percent of the ad volume, while the top of the funnel accounts for 60 percent.

I'm just giving an example here. But it's a good scenario, and I think your percentages are reasonable under normal circumstances. But what about during Black Week? That's when this happens.

Ahead of Black Week, you're feeding in plenty of traffic from the top of the funnel down. During Black Week, you turn the funnel upside down and focus on people who already know the brand. A large share of the communication and spend goes to existing customers. Why?

When you do that in Meta, shouldn't you exclude customers who've shopped within a certain period? No, absolutely not. I generally never exclude anyone. Even if you bought something the day before yesterday, you'll still see the ads one way or another, and they might annoy you. I don't exclude anyone, even if they bought the day before yesterday and the offer launches today.

I'm thinking you might want to split it into different campaigns. You want new-customer acquisition, but also returning customers. Meta will always go for the easiest customer, after all.

Yes, and that can mean you end up buying back customers who were already planning to shop. Absolutely, and that's fine. That spend will still be small relative to the total. But why do we turn the funnel around? Well…

…during Black Week, people are bombarded with offers from every direction. If you don't know a brand, a 20 or 30 percent discount won't make you think: ‘Now I have to buy this.’ I often use Sonos as a comparison, because they didn't used to offer discounts. But take Apple today. Apple hardly offers any discounts. Even during Black Week, they have…

…a gift card with your purchase. Yes, and it's small in proportion to what you're spending. Say it's equivalent to 10 percent, but it's still something. I know their sales shoot up during these Black Week offers.

Apple usually gives you a gift card when you buy. Because you almost never get a discount at Apple, even the little you do get is something good, and you sit there waiting for it. The contrast is Dell, or some other electronics brand with constant discounts. When you get 20 percent off a new Dell computer, you don't know whether the discount is meaningful, because you don't know what to compare it with. You might even hesitate and think: ‘Fuck, are they ripping me off?’

I always tend to check Prisjakt or Pricerunner. Yes, but even there it's difficult, because they play around with the numbers there too. The computers come in fifteen different models depending on the supplier or retailer. It's the same with TV screens and robot vacuum cleaners. The electronics industry has learned that there are price…

…bots that compare prices, and they deliberately make it difficult for consumers. Ahead of last Black Week, Rickard Lyko went public and said: ‘Yes, it's true. We raised prices in October so we could offer a more attractive deal in November, because that's what the rest of the industry does.’

There was a storm of criticism, but at least he was honest. Yeah, I like that he stood by it openly. There's a lot of number manipulation. Once you have a promotion that isn't perceived as promotional trickery, we really do see sales shoot up.

You say we should reverse the numbers. If the goal is revenue of 100 million, how many ads do we need? We need to start creating them and have a strategy for the volume required. A common mistake I've seen is people saying: ‘Ads don't work.’ Okay, how many do you have? ‘Five.’ Well, of course you need creative variety in your ad library, whether you're spending 10,000, 20,000, 30,000, or 100,000 a day…

…or 500,000, and whether you're on your way to one million or 100 million in revenue. You need a creative library with different types of ads: carousels, still images, GIFs, influencers, unboxing, educational material, and storytelling. You need a broad variety. It could be ten or fifteen ads, as long as they're completely different. The hope is that you can at least run 50 to 100 ads in parallel in the account.

Whether you're small or large. Preferably more, but not too many more. Somewhere around 1.5 times your CPM… No, sorry, 1.5 times your CPA. What does it cost to acquire a customer? If you acquire a new customer for an average of 300 kronor, you should allocate a daily budget of 450 kronor per ad.

If you have ten ads, we're talking about 4,500 kronor in spend per day in the account. If you have a hundred ads, that's about 45,000 in daily budget. The difference during Black Week is that you can scale the spend down a little. We usually scale down and focus more on retention, meaning bringing back traffic that already knows us and maximizing profit during these weeks or days.

So now you're saying people should reduce their spend? Yes, the competition is brutal, and people are bombarded with offers. You can increase it if you see it's having an effect, but initially we often reduce it to maximize profit. Once things have reached saturation and the majority of the people we intended to communicate with have bought, following the 80/20 principle, we can increase spend again. That might be Monday or Tuesday during Black Week if we started on Wednesday or Thursday the week before.

How do you measure that? I look directly at sales in Shopify, which is what we use in our case and what most people are on. I compare sales with what we've spent on advertising.

I basically ignore ROAS figures and other metrics the different platforms give us. What have we spent in total across the platforms, and how much have we sold in total?

Do you look at CPM or CTR at all? Not at all. For those who don't know: those are cost per thousand impressions and the proportion of people who click out of those who see the ad, respectively.

You become fairly immune to a CPM figure when it can vary by several hundred percent even within one ad account. It varies between different accounts too. Of course, I look at CPM over time. If it's a thousand percent more expensive than in October, I'm not going to spend as much in November, but usually it isn't that much more expensive. In Sweden, it can be 30, 50, or 70 percent more expensive during Black Week.

In the US, it can be several hundred percent more expensive. Either way, reaching people costs more. But when you say you should turn the sales funnel upside down…

…what happens to the campaigns you already have active? Should these be new campaigns, or should you put these ads in the existing ones? Usually you don't have any promotions leading into Black Week, but you might have…

I mean, most people might already have an October campaign running. Yes, I hope not. Or what do you mean? Not focused on Black Week, just ordinary advertising that's running.

Absolutely. Hopefully you're not running that many promotions ahead of Black Week, especially if you listen to our podcast. No, I don't mean campaigns in that sense, just regular ads you run at different times of the year that have nothing to do with Black Week.

My question for the listeners is: should they put Black Week ads into the existing campaigns that are already running? Ah, you mean the campaigns on Facebook?

Yes, exactly. Yes, you can do that. You can put your Black Week ads into existing ad sets and the existing campaign structure. Alternatively, you can create new ones. Then you'll most likely reach more of your returning customers and fewer new ones. That might be good. You want to give Facebook some kind of fresh start. It depends on the structure: how much were you spending, and how much revenue were you generating before?

There's a lot of judgment involved for a media buyer in this case, or you can ask Claude. Often it's better to use the existing structure than to create an entirely new one for Black Week ads.

So you keep both campaigns and ad sets? Yes, and you can pause, turn off, or rework ads. These days you can even change the copy in some ads and still retain the engagement. If you have a promotional ad that's worked well, keep running it. Just change the text so that Black Week is really obvious, and it'll still work.

Black Week is about making ads absolutely unmistakable. People have very short attention spans. They shouldn't have to read a whole piece of copy to find the offer at the bottom. It should be crystal clear in the image and in the text. The offer is visible and clear. Then we get good results from the ads and the advertising overall. It's no more complicated than that.

We have two more points before we move on. Or one left, actually: have you stress-tested your checkout and the system overall? What do I mean by that? Whether you have Shopify or something else, I'd test whatever is going live during Black Week. Different discount codes will be activated, and loads of emails will be sent. Test the links and the discount codes. Check whether the checkout can keep up when it gets a few hundred more visitors than normal. Can it handle ten times as many visitors, or does it crash? If you're on [unclear in the source: ett ask] or another system and have poor hosting so the site crashes, make sure to…

…stagger the emails so everyone doesn't arrive at once. When you stagger them, you also reduce the risk of ending up in the spam folder. That might be a side effect.

That's how it used to be, at least. I'm not sure whether it still is. I find it hard to see, but…

…now you can register for DMARC and similar things. They're not certificates, but things that show who the sender is. No, I'm not sure that works. I can imagine it worked better in the past, when the same mass email went out to hundreds of thousands of people.

No, fuck it. Let's move on. The test is making sure everything you've prepared and worked hard on is actually live, active, and working. That the backend is where it should be. That's easier said than done and takes hours. One option today is to put OpenAI or Claude to work stress-testing the site and going through the plan. As long as the links and discount codes are clear and you have [unclear in the source: enabard] or something similar, it can read through and check: does it work, and does the right time appear to be scheduled?

If it's supposed to go live at 8 p.m. on the Thursday evening the week before: do the site and discount code go live as intended? It can stress-test a lot for you. Then many Shopify Plus merchants can use SimGym, Shopify's simulation, to simulate…

…what it would look like if thousands of customers landed on the site at the same time. I actually looked this up: it's called email throttling. It produces fewer sudden volume spikes, better control, and the ability to prioritize engaged customers. But an important part is what kind of…

…sender reputation you have. The quality of the list and how recipients respond to the emails are more decisive than how you split up the send. So deliverability isn't affected that much by throttling. But it is affected to some extent…

Oh, it is, it says here, even today. That was interesting. Good point to add; I didn't know that. We usually send the emails at specific times. One example that came up was that you put your most important customers first. I'm thinking of Såpa, one of the brands we've mentioned.

They get to shop at the discounted price before everyone else. Yes, we do that with most brands. We usually close the site so only people with the password can get in.

But then you're letting the engaged visitors in first, so you are doing this. Yes, we prioritize people who've shopped with us before, more frequently, in greater volume, or over the most years. We do that. Maybe we're throttling indirectly.

I hadn't thought of it that way. Another positive thing is that we collect loads of email addresses for next year's Black Week, because people who can't access the site…

…get a little annoyed, but they understand. They think it's logical: the companies prioritize their actual customers, not just the new ones everyone is trying to chase and give the best offers to.

With some companies, people might feel it's fake. What's ‘scarcity’ in Swedish? Oh, demand, maybe? Limitation, then.

Yes, limitation. An artificial limitation. But as a company, you don't have an infinite number of products, so there is a real limitation. It's the same if you're a personal trainer at a gym and say: ‘I only have room for fifteen more clients’…

…because the person has an actual limit on their time. With Västerbottenssåpa, you now have the advantage of having built up enough demand that you actually risk selling out.

Yes, we have actually sold out many times, especially during Black Week. Now let's move on. I sincerely hope you listening have had time to plan, get to grips with this, and already have a large lead list. I also hope that, besides it being a lot of work, you think this is one of the most enjoyable times of the year. I do. Going to work during Black Week is one of the best feelings in the world. You see the numbers, revenue increases, and the copy falls into place.

The profitability is there. A lot of what you've been working toward. What do you recommend for people on the other side, who are doing badly? Shut down the store! No, I'm joking. Plan better for next year. Speaking of planning: this episode is also going to be about forecasting for next year.

We talk a lot about forecasting at the end of the year, because we're often involved in those discussions. Right now, for example, we're doing it at Västerbottenssåpa and Mockberg, and doing a lot of forward planning. What does a good budget for 2027 need to contain?

Tell us. If you want to grow from 30 to 50, or from 30 to 60, where do you start? Should you budget revenue first, or start with the desired EBT, profit multiple, or however you're thinking about it?

It's not a straightforward question. We usually start from two different elements. One is the data-driven side, where the data guides the argument. Then you look at the growth rate, demand, and how big the market is. Are you selling in Sweden, Europe, or the world? How big is your available market?

How much of it can we capture with the demand we believe we can generate next year? Then there are market conditions. Are interest rates going up and purchasing power going down? You also have the lipstick effect: when interest rates rise, more people may want to buy cheaper products like makeup. An indirect lipstick effect.

There's a lot to consider when you put together a forecast. I understand why many Swedes find it difficult and complicated. It ends up as a guessing game: we stick a finger in the air, feel how hard the wind is blowing, and set next year's budget accordingly.

Some CEO announces a number without much confidence, and no one on the team really understands how we're going to do it. We have to break the goal down. What does it mean in terms of returning customers, new customers, and products sold per month, week, or even day? What does it mean in sessions? How many people need to visit the site? When you break it down that concretely, you quickly see what needs to be done in week one for the budget to be achievable in week 45, as we approach Black Week.

You also have to buy what you're going to sell. It depends on whether you have print on demand or something similar; then you print based on demand. But if you're selling physical products worth 100 million…

…the products have to be there. You have to tie up quite a lot of capital. One question to ask yourself is whether we can afford to take the risk. I've met entrepreneurs…

…who brought in investors, and the investors wanted them to put their foot on the gas. Step on the gas, step on the gas. That led them to take more risk than they otherwise would have, and then they went bankrupt.

It happens really often. These are companies that have had revenue of over 100 million. Yes, it's very common. I'm not surprised when you say that, because I see it happening too.

Or they struggle for a few years before getting the problem under control and selling off excess stock, because they couldn't grow into the size they'd geared up for. It's important to think about how the company is going to grow by 50 percent…

…without sacrificing profitability. If you're going to grow by 50 percent, you need at least 50 percent more products in stock than the previous year.

You might have to place the order today to receive it in six months. If you tie up capital in inventory today, you're sacrificing some of the marketing budget that's supposed to generate the growth.

Exactly, it's not particularly easy or a straight line. If you take out a loan, you add interest and other costs that increase expenses, which people don't always factor in. That's why I say it's difficult to grow by 50 or 100 percent without sacrificing profitability, unless you have a well-supported calculation showing that things get cheaper as you grow. That's rarely the case.

It could be that customer acquisition costs go down, or that you have extremely high retention for the product. People come back without you needing to reactivate them very much through advertising. A soap case, for example: a fantastic product that makes people who try it want to stay. I can give an example of a company I heard about recently. They sell a repeat-purchase product that you'd happily subscribe to because it's convenient. They had several people making different ad creatives for them.

They said: ‘We're going to test this for a week.’ But they switched it off after just five days with a fairly low budget. I can't say which company it was. So they stopped before the test was complete and hadn't built up any substantial amount of data. They just thought: ‘Shit, this costs money. The ads aren't profitable.’

When I went onto the site, I saw that it was a product you'd happily subscribe to or buy in multipacks. They had different flavors and so on. So I'm thinking…

…that you can also improve sales by looking at the site, not just the ads. If they'd had bundles, for example buy two or three and get a small discount or something extra, or even just shown that packs were available, that would have helped. When you click an ad, you often land on a single product and don't know the others exist. If you have a product people can subscribe to, make that an available button too. There's a simple plugin for that in Shopify. That way, an ad that wasn't profitable can, through an increase in the customer's lifetime value and…

…also the average order value, become profitable. Interesting observation. Don't just look at the ad and think the ads are doing badly. It doesn't have to be just the ad. It could be that…

…the ads weren't good enough in this case. They'd allocated too little budget to even be able to evaluate them. But the site needs optimizing too, so you aren't looking at just one thing or the other.

Fair enough. Back to forecasting and budgeting: a simple example is to look at the share of returning customers we have today. Say revenue is 50 million and we have 10 percent returning customers. Then we have roughly 5 million in revenue for free next year.

But to grow by 10 percent, from 50 to 55 million, we still have to chase 50 million in new-customer revenue. One tip when you forecast over several years: compare two companies. One retains customers and gets 5 million from them every year. The other has poor customer service, delivery experience, and product, so they constantly have to sell to new customers. The first company can gradually grow as the budget increases, because it always retains a base of existing customers. I've also seen companies where the customers don't come…

…back. The only way to grow twice as much next year is to spend even more on ads, because none of last year's customers are coming back. We also know that prices over time…

…go up on the platforms, and you've already acquired the easiest customers first. So when we budget, we usually look at how much returning-customer business we have ‘for free’ and how much new-customer business we have to chase next year to fill the gap between where we are and where we want to be.

You can set a budget somewhat by gut feeling: we're at 50 million and want to get to 70. Or you can have a data-driven budget that says growth is currently 10 percent year on year, returning-customer business looks a certain way, and the market is big enough for around 10 percent natural growth. With interest rates continuing to rise and some uncertainty, perhaps we land somewhere between 50, 55, and 60 million.

But you want to reach 70. Then you can work backward: what has to be true for us to get there? The data-driven budget says 55–60. How do we bridge the gap between 55 and 70 million? Where are those 15 million going to come from? Then perhaps I have to take some bigger risks.

Maybe I need to bring in an influencer who wasn't in the plan, or do something differently from last year. You always have to do something differently, but here it might mean something crazy or radically different.

It doesn't have to be. Maybe you just need to do more of what you're already doing. That could be the case, but it rarely is. A logical, data-driven calculation should already have that growth built into next year's forecast. If you say you want to reach a different level from the one you can already see you can reach, you need to think differently.

That's what I mean: bridge the gap between where you want to get to and where the data shows you'll get if you keep doing more of the same. I'm thinking, for example, if you make more ads with greater variety…

That should already be in the forecast. That's what I mean. In a growth calculation, you should be doing more of what you're already doing, perhaps a little better and a little sharper.

But if you're going to achieve something you can't already see in the calculation, there are two ways to look at it. Personally, I work with experience-driven forecasting. What does that mean? I have a data-driven forecast and I love data. But if you bring me a data-driven forecast, I may still ask you to kindly take it and throw it in the bin, because I want ambition. I want to aim higher.

I want to stretch further than the data-driven forecast shows. I want to beat the market and the budget, do something a little more different, and take a little more risk to achieve what I believe is possible. Not just what the data shows is possible.

It's difficult. Pragmatically, the data-driven model is the most logical one for most e-commerce companies. I'm not saying you should take my model based on gut feeling and experience, set a number, and then work backward. I might have the knowledge to do that. But now, with the help of Claude and ChatGPT and the new models…

…it's starting to become much easier to have an experience-driven forecasting model like mine. The first thing you need to do is structure the data: what does it cost, what are the purchase prices, and what does new-customer acquisition look like? You need clean, good data you can trust.

What does bad data look like, for those who don't know? The e-commerce platform might show one sales figure and the ERP system a completely different one. The return rate might differ between the systems. Maybe someone has forgotten an integration in the ERP system, or the integration is lagging and not sending the returns into the e-commerce system.

Then you get different data points in different systems. And I haven't even mentioned that Google Analytics might show a third figure, and Meta and Google Ads a fourth and fifth. It's much more difficult than simply looking at one system.

What do you do when all the systems say different things? You need to trace it back: why do they say different things, and how are the integrations set up? It's about infrastructure and technical integrations, and deserves an episode of its own. Here, we'll have to assume you have clean data.

With the data, you can now use a fairly long prompt. I wonder whether we should put it in the show notes. Maybe we can do that, or link to a site if it doesn't fit. I can link to an external page where we post it. We'll see.

You can then use that prompt to work with the data. You connect Shopify, which in our case is the platform and our single source of truth. Here are the sales, inventory, product costs, sessions, new customers, and returning customers. Shopify is good at separating those out. With that data, you can produce a very good data-driven forecast.

Very logical and fully structured. So you just take your prompt? Yes, put it in, and then you can stretch the ambition: with everything we know about the company, we want to achieve more than the market data says. What needs to be true for us to go not from 50 to 55, but from 50 to 70 million?

We want to stretch further. How do we bridge the 15-million gap? What has to be true, which boxes do we have to tick, and when? How do we go about it? Claude reworks the model so you suddenly reach 70 million. It's easy to manipulate numbers in Excel.

But you also have to carry out the activities. The model will show you in black and white which activities are needed if you do more of what you already do. However, it can never say that if you work with mega-influencer X, Y, or Z, you'll get those 15 million you wouldn't otherwise have got.

Or that a trend will suddenly emerge and produce a particular outcome. It can't predict that or make suggestions based on it. That gap still exists, but it's easier to bridge with AI. Another tip is to look at the products you sell using Google Trends.

If you sell certain products and want to grow but see search volume declining, that could be an indicator that you need to broaden the product range or find another angle in how you present it. Demand might still be there, but people have changed the words they use.

True. You have Google Trends on the one hand, and various SEO tools on the other. These days, SEO tools have built-in AI, so you can see what people are searching for and what's trending in the AI engines.

That has to be included in the calculation. If you're in a naturally growing market, you might grow by 10 percent year on year without changing anything, simply because demand is increasing. Alternatively, you're in a declining market.

Yes, and then in some cases it doesn't matter… It matters, it always matters! But take working in the newspaper industry as an example.

No, let's take Babyland, which has just gone bankrupt. They point to fewer babies being born, which means less demand for baby products.

The same thing happened many years ago, when investing in preschools was hot. Investors thought it would keep going up…

…because the population was growing and immigration was high. Yes, exactly. But they didn't account for it declining. When suddenly fewer children are born or…

…immigration decreases, you can't maintain the same positive numbers. What I mean is that you may need to evaluate whether the product category or the market itself makes the growth target harder. But even if you're Babyland in a stagnant market, or as in the baby world where fewer babies are being born…

…you can still take a share of the existing market, because children are still being born. Your market share ought to be able to grow. How do we do that when the market isn't growing? It's obviously harder than in a naturally growing market. A tailwind is easier than a headwind.

But it's not impossible. If I were Babyland, I might have restructured the company and focused more on high-margin products that are easy to ship and don't require as much storage space.

There are plenty of those, and loads of digital products that, as far as I know, they haven't invested in. From the outside, it was easy to see why they were hemorrhaging money. I think in their last financial year they had revenue of around 153 million and lost fifty-something million. That's why they had to throw in the towel. If I'd been involved in running Babyland, I'd have had different ideas about the structure and made radical changes…

…to survive in a market in crisis. I looked it up: revenue for the full year 2025 was 155.7 million, a decrease of 4 percent from 162.9 million. At the same time, the loss increased. Profit after financial items was minus 53.2 million, more than double the deficit of 19.5 million the year before. Yes, roughly what I said, give or take a million. It doesn't change the reasoning. Moving on: several things happen when we give AI access to Shopify, Meta Ads, Google Ads, stock levels, and purchase prices for forecasting. What previously required a fairly senior CFO, or a great deal of your time…

…can now be done with a long prompt, a few revisions, and a dialogue. Leave the computer working overnight and you can get very good results. You don't even use up your entire weekly usage allowance. It's efficient, if we're talking about the $100-a-month plan and the weekly usage limit.

You can do a lot. One thing AI can help prevent is cash flow problems as you grow. It can tell you how much inventory you need to buy now to reach a certain future revenue level, and whether it's possible. What risk are you taking, and what risk level are we at? A bit like investing in the stock market and choosing high-, low-, or medium-risk funds. We can project next year's revenue in the same way.

One tip is to ask it to do that on a scale, rather than just writing high or low. I recently read a study from before AI. When managers asked for feedback in terms of high or low risk and someone answered high risk…

…they then asked what percentage range that meant. Some people interpreted it as 80 percent, others as 20 percent.

A wide range, in other words. High, low, and in between are too abstract. A scale and an explanation of what it's based on are better. If someone says a particular score is good or bad, there has to be a scoring system…

…that explains why the risk is higher, with a certain number of points. True, that's incredibly good feedback. You need a scale to make what low and high risk mean concrete. AI can also hallucinate…

…which brings us to the next point: what do we do to prevent AI from hallucinating in the forecast? We can see that AI has got better at not making things up out of thin air.

You can also build quality checks into the prompt and have another AI fact-check it. I usually put two AIs to work talking to each other. I take the result from Claude, put it into OpenAI, and ask it to check that the numbers are correct. Not whether the assumptions or the forecast are good or bad. That's a broad range, and then it does an extensive job and almost always concludes that something is bad for some reason. You have to define what you mean.

It's dangerous just to say ‘make a good page.’ What does good mean to you? Nice animations? Is the deciding factor the price or how quickly it can be made? I also have a security agent that reviews what the other agent has done. One agent was downloading receipts for me.

The other agent went through the receipts and compared them with the transactions in the accounting software. Then it said: ‘These are made up.’ The agent that was supposed to download certain receipts hadn't been able to get into one of the platforms.

Oh, so it made up its own? Yes, it generated figures based on what it thought we'd bought and what VAT rate the product should have. It created its own receipts that looked like PDFs, but when you opened them, they weren't real receipts. The other AI agent, which was supposed to check them, could see that this was…

…made up. Fuck, that's dangerous! It said: ‘Oh, but I thought you wanted receipts.’ Yes, real receipts! I know OpenAI, and also [unclear in the source: kad], have put in a safeguard against generating receipts. You can get around the safeguard, but…

…it did this entirely on its own. I didn't want it to generate receipts. No, I know, it's completely insane. But if you write a prompt like: ‘Can you generate a receipt for a MacBook Air from Elgiganten [unclear in the source: 20217]?’…

…you can't do it. I've never tried. They've put in a safeguard. I heard about it on another podcast, then tested it myself, and it turned out to be true. That safeguard is there.

A strong reason to double-check is to avoid hallucinations. The other thing is making sure you put in the right numbers. Even if it double-checks, if it bases that on incorrect numbers like we discussed at the beginning, no agent will notice that the forecast is wrong. It's built on the wrong numbers. Correct numbers in, correct numbers out will really always be…

Yes, at least right now, the rule when it comes to working with AI. If you want to work with AI in the best way, make sure you put in accurate numbers. Then we're talking about the Facebook pixel, Google Tag Manager, and all the platforms that…

Above all, the ERP system and the e-commerce system need to work together, sync, and have the right data. Does it say you have 3,000 of a product in stock? Make sure the stock status is correct. I've experienced this myself: seeing ads and thinking, ‘I could see myself buying this.’ I click the ad and it just says ‘Sold out.’ Then I think…

Yes, I know. Now they're burning money on something unnecessary. That actually happens. You were planning to buy, but you go to a competitor and buy there instead.

Yes, because I still want certain products. But you can also use AI to set it up so that if a product sells out, it goes in and turns off the ad.

That's fucking slick. With that said: we also have a KPI scorecard that we work with at Porobic Group. I was planning to link that in the show notes too. That's why I'm saying there'll be quite a lot to link to. The best thing would be one link to a site where you can choose: do you want the prompt? Do you want the scorecard?

Do you want both? Then we'll send them to you completely free, because sharing is caring. Now we're going into the ad of the week, because we're also approaching the end of the episode.

I wouldn't say the end of the episode. The ad of the week usually takes a little while. The highlight of the week's episode! I'm just kidding. This is actually starting to become the highlight of the week for real. I know lots of people write in wanting their ad dissected. But Fredrik has found a company called Danger Coffee.

If you didn't see last week, we dissected Eight Sleep then. It almost immediately went viral around the brand. It was fun to see so many people engage with some of the things we misspoke about, and probably some interesting angles we didn't bring up. But this week we're covering Danger Coffee. Alright. It's a bit like: beware of this coffee. Is it too good or too bad?

He was at Superhuman Summit this year too. A very distinctive character. But let's look at one of the ads here. We'll take the first one. I like the tagline: ‘Wait, is that Danger Coffee on Erling Haaland's counter? Small things to some, not to us. Thank you, Erling. Go watch the full video on his channel.’ Okay, that was a good call to action to go to another channel.

Let's click it. It's on his YouTube channel. ‘You want me to make you one? Coffee is a superfood in my eyes, if you do it right.’ ‘Why?’ ‘Depends on the quality, when you drink it and, of course, how. So yeah, I'm into my coffee. Here you go. Try, see if you like it. This is how I drink it. With milk, maple syrup, to protect the caffeine with fat and sugar.’

‘Really? I can make you one.’ Okay. I don't know what I think of the ad. I like that they have one of the football world's, and now one of our era's, great icons as an ambassador. But I don't know whether I like the ad itself. ‘If you want, I can make you one’ — what kind of hook is that? Since I've seen the original video too: this isn't really an ad in itself. He's talking about this for free in his own vlog. He's just vlogging: ‘This is my morning routine.’

Then they've gone: ‘Ah, he has it in his morning routine,’ cut out a section of the vlog for an ad, and presumably asked him: ‘Can we run this as an ad?’

I assume so, otherwise they're risking… That's why they have the tagline: ‘Wait, what's Danger Coffee doing in Erling Haaland's…’ Then he talks a lot about maple and fat. It feels more like a sugary drink than a morning coffee with fat. I don't know. I don't know what I think of the ad for Danger Coffee.

I don't get any sense of Danger Coffee. I don't know what it is, and I don't really see anything. What I can imagine is that this is a middle-of-funnel ad for people who've become aware of Danger Coffee.

Then they get this to create the association: ‘If Erling drinks Danger Coffee, maybe there's something to it.’ Because he's one of the best football players of our time.

Fine. I want to be a little more like Erling Haaland too. Nice. Let's hope he has Västerbottenssåpa in his next vlog. Yes, exactly. Okay, next one then. Don't take this iced coffee one. Let's take one that's a little more… That one: ‘Your body runs on minerals.’ Let's take that, because it's their own…

Or ‘Why put butter in your coffee?’ Let's take one of those ads. Okay, here we go: ‘Your body runs on minerals. The problem is most of us are not getting enough of the good stuff. Danger Coffee makes it easy with 50 trace minerals built right into your morning cup. No weird taste, just clean, remineralized coffee you can feel.’ And various hashtags.

Let's try playing the ad. Danger Coffee ads. ‘Your body runs on minerals. Magnesium, potassium, calcium and over…’ What does he say? ‘Danger Coffee, 50-plus minerals added’ is the context box, as I call it, that's in there. Okay, let's play it then.

‘Your body runs on minerals. Magnesium, potassium, calcium and over 50 trace minerals that you're probably not getting because the soil is depleted and the water is filtered. So I built a coffee that puts minerals back in. Danger Coffee is mold-free. It's third-party lab-tested, and it's remineralized with over 50 trace electrolyte minerals. You can't taste the minerals, but you'll feel the difference. No jitters, no crash. An energy that lasts until dinner.’

I like it. This is a much better ad, because you find out what it is I'm buying. What is Danger Coffee? He starts with a strong statement: ‘Your body runs on minerals.’ As if it's an established fact: it's really important for your body to get minerals. Yes, it is.

He doesn't go into depth about why it is. That's probably somewhere on the site, I hope. But the whole USP of the coffee seems to be that it's like ordinary coffee, any coffee, but I've put in some extra minerals because the soil has fewer minerals being produced…

Or, as he said, if you filter the water. In unfiltered, ordinary drinking water, you usually get minerals, but you also get a lot of junk. That's why people choose to filter their water. But then…

How are you going to get your minerals? That's where we have Danger Coffee. I still like the name. But then he says: coffee that lasts all day. ‘No jitters, no crash.’ For it to do that, it isn't the minerals doing it. Then it has to be loaded with something else. Probably fat.

Then they have the other ad, ‘Why put butter in your coffee?’, because that doesn't come across at all in the first ad. No, exactly. ‘Why put butter in your coffee? In case you missed it: Danger Decaf is packed with polyphenols that feed your gut bacteria and are good for your mitochondria.’ This really is for serious biohackers.

Yes. ‘Learn more in this discussion with Dave Asprey and Dr. Gundry.’ Dr. Gundry's podcast. Okay, let's play it. It's a podcast ad. He's appeared on a podcast, one of the founders of Danger Coffee.

Here it just says, as the hook — or not the hook, but the call to action — ‘Danger Coffee, food and drink.’ That's fine. ‘Like page’ is the call to action. That's a little annoying. Okay, so this one is trying to drive likes, I assume, instead of…

‘It turns out that fat enhances the absorption of polyphenols. If you were to, I don't know, take a shot of olive oil, which has its own polyphenols, and then drink your coffee, or you were to blend fat in your coffee, the good fats, magically you're going to get even more benefits from your coffee. And one of the reasons is polyphenol absorption.’

‘Another one is the pharmacokinetics of caffeine. What happens when you blend it with fat, or even when your body does it, is it changes the structure of the water that is in coffee so that your body can more quickly use it inside cells. This was validated by Dr. Gerald Pollack at the University of Washington. I did fund that study, but it was a basic biochemistry study, not a test of my coffee.’

‘We tested MCT oil and ghee and found that they made a very large exclusion zone when you put water up against them, larger than he's seen from anything else. So what I think is happening when you have fat with your coffee is that, in the gut, as it gets mixed up, you get some of that. If you just blend it or add some to it, you get benefits from it. It turns out that fat enhances the absorption of polyphenols.’

I don't know what I think of this. First of all, it's a good hook: ‘It turns out that fat enhances…’ He essentially has a new theory.

For many people, this will be fairly new: ‘Oh, it seems that consuming fat increases the ability to absorb polyphenols,’ which are what strengthen a lot of things in the body, inflammation, anti-inflammation, and so on. It's a good statement to have, because it makes you want to see more.

However, I don't like how constantly the ad, or the clip, is cut. Almost every word has a new cut. Then I think: is this fake or real? Couldn't he say a whole sentence with a straight face without needing to cut?

If we can sit here and record a whole podcast almost without cuts — perhaps we have a slip of the tongue that we cut out — why can't he nail one or two sentences…

…when it's supposed to be an ad? I don't know, but it bothers me. Then he says he's done a study, or that it's been proven in a study. He doesn't say how many people took part. He doesn't say it's an established, large study. Has it been verified? But he mentions that he helped contribute…

I think that's honest, at least. Yes, with money so the study could happen. That's fine and very honest. But then I become skeptical about whether what he said before is true. Did you really need a study that you fund in order to sell your coffee?

That's what's interesting to look at. It's the same if you say there are studies showing certain things aren't dangerous, but then they're funded by the companies that benefit from selling the products.

But I haven't read the study, so I can't say anything. What do we think of the ads? The first one, with Erling Haaland, wasn't very good at all. It might have its place in the ad account, but as an ad, if you cherry-pick it out of the account, it's worthless. About 3.5 out of 10.

The second one is a little better. Still not a good enough ad. Maybe 6.5 at most. The podcast ad, maybe a five out of ten. None of the ads gives you a complete picture of what Danger Coffee is. Maybe the whole idea is to spark curiosity so you go to the site and read more. But a good ad should be able to sell the product within the ad, so the mental conversion has already happened. When you reach the site, you have an easy route to buying, instead of needing to read up on loads of things: ‘Where do I find this?’

A tired consumer won't want to do that homework. Make it easy for the customer. People have so much going on around them, ads and so on. That's it. We don't need to look at the entire ad library; it'll take…

No, exactly, a very long time. But if you're interested in more ads, you can go in. Exactly, go into their ad account and have a look if you're still interested in Danger Coffee. In any case, the name feels very different.

What do you think? You're also interested in coffee and in it being, as they write there, free of mold. I'm not interested. I don't feel tempted to buy Danger Coffee. The name alone makes me think: ‘Wow, what am I putting into my body?’

The fact that they say ‘mold-free’ doesn't say much. Most coffee, at least in Sweden, is mold-free. Especially when it's dry-roasted, vacuum-packed, and then arrives at your home. Of course there are mold spores or something in tiny amounts that aren't really particularly dangerous for you…

…otherwise it would have been… But the beneficial effects of coffee are well studied and well documented. Does any of these ads, or all three together, make me buy Danger Coffee? Not even close. I wouldn't even look at the site, if I'm completely honest.

I do want to give a tip here, because we're talking about coffee having positive effects. I have someone close to me whose anxiety decreased enormously when the person gave up coffee.

Fine. Everything is individual; we know that with everything you put on and into your body. The funny thing was that she was having lunch with a colleague who asked: ‘Aren't you having a cinnamon bun and a coffee?’ She said: ‘No, unfortunately, I've realized that…’

‘…coffee is bad for me. It gives me anxiety.’ Her colleague said: ‘No, but it's fine. I just take antidepressants, and then it's all good.’ What? Okay. The realization eventually came from people asking: ‘Did you eat sugar before you got anxious? Did you drink coffee before you got anxious?’

She tried gradually excluding those things and realized: ‘Shit, that was a major contributing factor.’ That's why I want to bring it up. Not everyone is sensitive. I don't get anxiety from drinking coffee, but some people actually do.

Caffeine intolerance exists. That's why one of their ads, from what I saw, also focused on decaf. That's fine. With that said, that's a wrap for the ad of the week and this week's episode. If you've listened this far…

…we really want to encourage you to share the podcast with friends and colleagues in the industry who you know will benefit from the forecasting and Black Week, which we're in right now. Click the link and download the materials.

Click the link in the show notes and download everything we've encouraged you to get and are generously sharing. You have nothing to lose. In the worst-case scenario, you get a good, good-enough forecast.

Thanks so much for this week's episode and for spending this time with us. We'll see you again next Thursday, at the same time as usual. Bye! Yeah.