Stampen Media Programmatic

Håkan Hamrin, Head of Programmatic

February 12, 2025 Case no.19

Growing online revenue

How Stampen Media is maximising programmatic display revenue

In five years Stampen Media, a local news group based on the west coast of Sweden, has grown their programmatic display business from 3 million USD/year to 7 million USD/year. And this has been done while the group has been focusing on reader revenue which makes it astonishing. This webinar will cover strategies to maximise revenue from national advertisers while building the local online display market. Learn about header bidding, inventory-optimisation, CPM-pricing and on how to open the paywall for ads.

Presented by Håkan Hamrin, Head of Programmatic at Stampen Media (until Jan 31 2025)

Heading 4

By Niklas Jonason

Article Summary

Stampen Media, a local news group in Sweden, has successfully grown its programmatic display revenue from 3 million USD/year to 7 million USD/year over a five-year period. This growth has been achieved while the group has been focusing on reader revenue, which is quite remarkable.

The key factors behind this success are server-side header bidding, combining more ads with a retained user experience and high viewability, and experimenting with different shares of articles behind paywalls.

The first factor, server-side header bidding, allows Stampen to add more demand sources to its programmatic advertising, which helps increase competition and drive up prices. The second factor involves using AI to optimize ad placements without harming the user experience. The third factor involves strategically placing articles behind (and sometimes in front of) paywalls to balance the needs of both advertising and consumer revenue.

Background on Stampen Local Media

Stampen is one of the leading local news media groups in Sweden, focusing on 17 news brands, out of which most are paid for, in the western part of the country. The group has 1500 employees and has its headquarters is in Gothenburg, the second biggest metropolitan area in the country. The main news provider there is "Göteborgs-Posten" which is the main title of the group. The Stampen Group is owned by Polaris in Norway together with two Swedish local news media groups. The main owner of Polaris is the Norwegian Schibsted group.

A quick pitch about the speaker and his case

A brief presentation of the man behind the case: Håkan Hamrin has 15 years years of experience working with programmatic advertising within Swedish media groups, the last five years for the Stampen Media Group, based on the west coast of Sweden. Håkan however has worked out of his office in the northern town of Östersund.

The short pitch of what Håkan and his colleagues at Stampen has achieved can be summarized in the revenue results, shown below, going from 3 million USD/year to 7 million USD/year in revenue from 2020 (the green circle) to today. Håkan emphasize three important factors behind this progress:

Server-side header bidding, which makes it possible to add more demand.
A solution to combine more ads with a retained user experience and high viewability
Experimenting with different shares of articles behind paywalls to meet the demand from the advertising side without harming the consumer business

First a brief introduction for beginners in programmatic advertising

Before going into detail we asked our speaker Håkan Hamrin to start with a simple primer on the topic of programmatic advertising. It is an area with abbreviations and funny words that could need some explanations (that you may have noticed above).

First let us describe what programmatic advertising is about. The traditional method of buying and selling online advertising involves manual negotiations, contracts and handling of material and copy flows between advertisers and publishers. Programmatic advertising, on the other hand, is the automated buying and selling of online advertising space. It uses technology to target specific audiences and optimize ad campaigns in real time. This type of advertising is more efficient and data-driven than traditional methods. Håkan estimates that 80-90% of the national Swedish online advertising is programmatic today. And of course the buyers have been market drivers. Instead of sending a lot of material manually and getting invoices from a lot of publishers, buyers can just work on one platform and get all the economical flow in one place and in addition to get access to targeting.

In programmatic advertising, the advertiser works in the demand-side platform, the DSP, where they can put their campaigns and the material and define the targeting and so on. And on the other side, we have the publishers, which are working with the supply-side platform, the SSPs. And in the SSP, they offer a lot of ad placements, which are for sale. See the illustration below.

In the ad exchange, the DSP can buy ad impressions from the SSP. And the SSP will pick, most of the time at least, the highest bid from the DSP and then show that ad on the publisher's website. To understand one of the biggest secrets behind the success of this case we need to understand the difference between “waterfall auction” and “header bidding”:

In the illustration below you can see the “waterfall auction” on the left. We have a floor price, which is the lowest price accepted by the publishers of $ 2.1. The first partner has a bid of $ 1.7, which is below the floor. So then in the waterfall, the request is sent to the next partner, partner number two and they have a bid on $ 2.15 which is above the floor price, and will therefore be the winning bid. If that had been below the floor price as well, the request would then have been sent to the partner three. And as you can see, partner three had a higher bid than partner two. However since the request came to partner two before, they had the winning bid.
With the “header bidding” on the right side, the requests are sent to all the partners at the same time. And then you can choose the partner that has the highest bid, which then is $ 2.5. So header bidding not only has made it possible for the publisher to optimize the revenue on every single ad impression, it has also made it possible for the publisher to add more partners working with more bidders to raise the demand to a higher level for them.

And speaking about header bidding, the next illustration shows how it works in detail. The client-side auction is made in the browser of the user. But there's also a server-side auction, which is done on an external server and doesn't affect the browser of the user. So in the server-side auction, you can have a lot of bidders because it doesn't affect the page in any way. On the client-side, though, you can have a few bidders. Otherwise, it will affect the page you're on and will make it slower. So you can have a lot of bidders on the server-side, and then the highest bid will go into the client-side auction and be compared with the bids from the client-side auction. And then the highest bid there will be sent to the wrapper, in our case, “Livewrapper”. The wrapper will then have a communication with the ad server to see if the ad server has any ads that should be chosen instead. And the winning ad is decided by the ad server, depending on the rules set in the ad server. And then the ad is shown on the website in the user browser. So this makes it possible to have a lot of bidders, but also to combine it with those ads you have in the ad server, the direct or locally sold campaigns.

The three main businesses of a newspaper web site

For Håkan there are, at least, three different businesses of a newspaper web site:

A consumer business based on subscriptions building on a paywall, where you want to get as many people as possible to convert to paying customers.
Programmatic national advertising
Local and/or direct advertising

The astonishing trick of Håkan and his team is, of course, that the impressive growth of national ad revenue has been achieved while the group has been focusing on closing the web sites with paywalls to concentrate on reader revenue. And at the same time they are building the local online display market.

From that perspective of increasing reader revenue, you will, of course, put a lot of articles behind paywalls to drive the subscriptions. However you also have the advertising side, which wants to have a lot of ad impressions, and is probably less keen on putting content behind the paywalls, because that will actually lower the number of ad impressions.

On the advertising side, you have the programmatic national advertising, which is kind of automated and sometimes has lower CPMs, but sometimes has really good CPMs, as well.

Then, last but not least, you have the local direct advertising, which often is sold on impressions. You give the customer a number of impressions, and they will pay you based on those number of impressions. But sometimes the local or direct sales are based on a share of voice business model as well, where you get, for example, 20% of the impressions on the site for that day or that week.

And this can make it quite tricky to optimize all those three businesses at the same time. And that has been a challenge for Stampin as well. So this is where the trick comes in. In the rest of this article Håkan shares how it has been done!

The main success factor: Adding demand thanks to server-side header bidding

Let's start with the first success factor; server-side header bidding, which makes it possible to add more demand. We have described the difference between the server-side and the client-side header bidding. This is actually based on classical economical theory that says that where you have a certain supply, and you add more demand, the price will increase. This is the simple theory that Stampen has been working on through adding a lot of bidders to increase demand and therefore prices and revenue in its programmatic business.

The next diagram below shows the number of bidders (in black) connected to Stampen's header bidding wrapper related to total revenue from national programmatic advertising (in orange). Stampen has gone from about 20 to almost 90 bidders. Håkan remarks: "It is clear that we have a close connection between the number of bidders and the revenue. So adding more bidders has been quite good for us to add more bidders. And we have seen other publishers doing this as well."

The second success factor: combining more ads with a retained user experience and high viewability

Let us move on to the second success factor: a solution to combine more ads with a retained user experience and high viewability.

If you have a consumer business and you want to drive subscriptions, there is a risk that you will lose some advertising inventory. Therefore you need to do some stuff to add more inventory based on these requirements. Stampen has been working with the company Browsi and its AI driven inventory solution. Browsi collects a lot of data points from the user and from that data they try to predict the user behavior and then propose the best placements for ads while not harming the user experience. So the solution combines the use of first party data (data from visits to Stampens own properties) with AI.

In the illustration below we can see some examples on how it can work:

The first example (on the very left) is the fast scroller. You normally show an ad quite high up on the page. Instead, you need to load the ad further down.
In the second example, social, we have a more social scroller, as to say, which is scrolling a bit slower. For this type of user you have the time to load ads higher up on the page. The ads will still be viewable while not harming the user experience.
If the user has a slow internet connection as in the third example, you need to wait with loading the ads until there is enough internet bandwidth to actually load them. So then you will have to load them further down as well.
And of course, in the fourth example, if there's a non-human bot, then we won't load ads at all because we don't want to show ads for bots.

The following slides below show the results when using these measures mentioned above.

The first one shows the UX metrics: “the average scroll depth” of Stampen web sites. The blue line marks where we started using Browsi and as you can see, there's been a small decline in the scroll depth at first, but basically, after that, has been at the same level.

If we look at the second UX metrics below which is “average time on page”, we actually had a bit increased average time on page after implementing Browsy on our sites.

So this tells us that for the user, it has been pretty okay to have more ads. It hasn't affected the user behavior at least. But let's also have a look at how revenue has been affected.

So this tells us that for the user, it has been pretty okay to have more ads. It hasn't affected the user behavior at least.

Revenue development in detail

Let's also have a look at the revenue and how that has been affected.

So the prerequisites that results in increased revenue are:

Stampen went from 1.8 ads per page view to 3.3 ads per page view.
If you make the assumption that we'll have 100 million page views, we have a 75% fill rate (“Fill rate” is how large a share of the inventory that is actually sold). It's quite hard to sell 100% of the inventory, but 75%, is quite okay.
The average price measured in CPM is 35 SEK (about € 3). CPM means the price per 1000 ad impressions or to be exact Cost Per Mille.

Håkan sais: "With these parameters revenue has gone from € 4.7 million up to € 8.6 million. So with this we can conclude that it is good for the advertising revenue and for business to implement this solution. And if we look at the third factor, experimenting with different shares of article behind paywalls to meet the demand from advertising side without harming the consumer business, was the next challenge."

The third success factor: Experimenting to meet advertising demand without harming the consumer business

Håkan prroves his case: "We have noticed that we have higher CPMs around salary time which is around the 25th each month in Sweden. And of course, when people get money, they will buy things. So a lot of advertisers want to reach the people around that time of the month because they know that their sales will be higher". In the diagram below, you can clearly see that the CPMs are higher around those dates.

In a similar diagram below it is shown that Stampen also added revenue which is also depending on the success factors mentioned above of course.

And as you can see here, we have revenue tops as well around those salary dates. "Those revenue tops weren't there before because we didn't have the supply to meet all the demand around those weeks. But we started to put less articles behind the paywalls during those payment weeks.

We did a couple of tests first to see if advertising harms the consumer business, but we didn't see any lower share in the conversions to subscribers. The newsroom also thought that it was a good idea to bring more people into the pool they could fish from, as to say, they had more people reading their articles and perhaps they could be a converted reader later on."

So to summarize it, the three most important factors for the programmatic revenue raise we've had those last years, server-side header bidding, which makes it possible to add more demand, a lot of demand, a solution to combine more ads with a retained user experience and high viability, experimenting with different shares of article behind paywalls to meet the demand from advertising side without harming the consumer business.

Q&A with Håkan Hamrin

I recently visited a local newspaper group up in the north of Sweden, and they use similar methods and platforms like you do Håkan. They mentioned that they had developed something they call the “article honk” which works like this: when they have a story which is behind paywall that seems to be getting attention, there is a “honk” and an editor may let it free for everyone to read with the goal of getting ad impressions and with that higher ad revenue. Tell us about that choice and when that decision is being made and what revenue that can bring.

“Yes we have looked at the “article honk” as well from Stamper's side. We haven't built a solution for it. We have more of a manual process where we have one guy looking at the statistics. And when he or she sees that some article is going up in the statistics, then he or she can tell the editorial people that, hey, here's an article going away in traffic. Maybe you want to put it in front of the paywall instead of behind the paywall. But I think that that's kind of the same logic as we have used for the things we have done when it is salary time every month. You want to experiment with the share of articles you put behind or in front of the paywall and see what can bring the most value to the table. Is it more worth it to have it behind the paywall and drive subscription? Or is it more worth to have an article not behind paywalls and drive a lot of ad impressions? So that's not easy work to do. But I think you need to look at some different ways to optimize that."

You mentioned that local online display advertising has more traditional business models, like buying per share of voice or per CPM. Is local advertising also moving to programmatic?

"I think that depends on the market a bit. I know some places in Sweden, there are local agencies that are working with the programmatic business and help advertisers to advertise programmatically. But I think for the small advertiser, it's not that easy to start working with a DSP and start to do this yourself. You want to have some help with it. I think there are also a lot of media companies, Stampin Media included, that can help the advertiser with the programmatic advertising as well, which I think is quite a good idea. I mean, as a local newspaper, you want to be a bit like an agency and help the advertiser reach out where they really need to reach out.
And about the share of voice and CPMs. I think that for a very small site, it can be a really good idea to have the share of voice because then you can get better paid from the advertiser. But when you are growing the traffic and advertisers will get a lot more impressions for the same price that they had when you had low traffic, that's not a good business. And when you have quite a lot of traffic and you need to optimize between the programmatic and the local advertising, you want to work with the CPM because the CPM level will be what you compare between those different businesses. And to optimize the revenue, you always want to pick the ad that has the highest CPM. I think that's quite an important thing to do to raise the programmatic revenue as well."

You mentioned the CPM 35 kr (€3). How does that compare
to locally sold ads?

"I would say that the average CPM is lower than the local. Yeah, definitely. When you can sell it direct or locally, you can get higher CPMs, but most of the publishers will have a problem with the fill rate. You can't have, you know, the fill rate and the share of the ad impressions that you actually sell. A lot of publishers have problems filling 100% and if you fill it 100% only from local advertisers, then most probably you won't have very good CPM levels.
You need a lot of demand to have a high fill rate combined with high CPMs as well. One more thing, I mean, the CPM 35 (€3), it's not that high CPM, but there's a lot of ad impressions as well and you have a lot of impressions that has like 60 (€6) or 70 (€7) CPM, but you know, you have a lot of the bottom as well, which doesn't have that high CPMs. You might have it on 10 (€1) or 15 (€1,5).
So the average will be in the middle somewhere. ”

When I looked at this, maybe 18 months ago, there were other solutions used in continental Europe, where publishers had deals with an advertising supplier that payed a fixed amount for each full web page on which this supplier put in ads and let the publisher cover the costs, so to say. So maybe your solution where you optimize the revenue is more intelligent. And if yes can the publisher earn more money on fewer ads?

"I can understand working with one supplier, because then you'll normally get a high fill rate. But I am very sure that if you're working with several suppliers, instead of just one bidder, you will get higher CPMs. You will get a better result instead.

It is another discussion, but yeah, it might be that if you get a good CPM level on the placements, you can actually lower the number of ad placements you have.”

Have you explored programmatic in direct sales? How does that stack up in CPMs?

"We have been working with the guaranteed direct deals, as may be the case locally, and programmatic guaranteed deals as well. And yes, CPMs are higher. So that has been a good thing for us as well. Although you can sell a bit, perhaps 20-25% through that place, you still have to fill the rest of the impression. But yeah, definitely a good way to raise the CPMs to work with that. How would you fit in?"

I guess some media organizations still have annual contracts. Have you experienced that? The sales department says, well, this guy has an annual contract, so you need to put him into the system. And how do you treat those customers?

"I think that different kinds of discounts aren't that unusual in the media business, correct? Well, I wouldn't say it's a big problem. The most important part, I would say, is to use the CPM you actually get from the customers. I mean, if you have a price of 100 CPM, and then they get 70% discount, and they only work paying 30 CPM, then you have to use 30 CPM in the optimization instead of 100. Because then you will optimize the revenue, and then it will be a good business. And they will get a lot of impressions on 30 as well. But if you're using 100, you're sub-optimizing, and that's never good for the revenue, at least. "

Useful links and contact information

The webinar presentation can be downloaded here.
The presenter, Håkan Hamrin, was Head of programmatic at Stampen Media group until the end of January this year. Since February he is Customer Success Manager at the Swedish company Livewrapped that has the mission the mission to help publishers get more revenue. His new employer offers a header bidding wrapper, which includes both a client-side and a server-side wrapper. Håkans goal is to help publishers “do what we did at Stampen”. He can be reached on his mobile number +46 703 161 697 or via his e-mail-address hakan@livewrapped.com
Link to Stampen media and som of the news properties: www.gp.se (metropolitan), www.hallandsposten.se (medium sized) and www.sttidningen.se (small)
The supplier Browsi was also mentioned in the presentation and in the article above.
This is the second case from Stampen Media in the program. Check the first from August 2024 about e-paper that can be found here.

You are welcome to contact the WAN-IFRA Innovate Local team, if you have questions or examples of similar cases.
Cecilia Campbell: c.campbell@wan-ifra.org
Niklas Jonason: n.jonason@wan-ifra.org