Digital transformation — The Mittmedia way: User loyalty is our business
Mittmedia is a publisher of regional and local news brands in Sweden.
We have the same challenges as most publishers do. Print revenue in decline and an organization filled with tradition.
But we also have a culture of innovation and a team that’s determined to win.
I want to share with you how we, with the help of data and focus on the users, have built real digital transformation.
It looks impressive when you look at Mittmedia’s regions on a map like this. We cover a third of Sweden’s territory with local journalism. We are the main local news source for the 1,7 million people living in our regions. We take our mission very seriously.
We are owned by a trust but we must make a profit. Not for the owner’s sake. But for the readers. We invest our profit in making our products better. Because that is our mission and reason to be: to provide high-quality local journalism.
We are known to our readers as strong local newspapers. Some with an almost 200-year-old history.
We have today an expanding digital business on the foundation of the loyal audience. But we struggled for some time with our digital business strategy.
Digital reader revenue — from niche revenue to the main focus
This is a simplified image of how digital news products generate revenue:
We get traffic to our news sites that give us ad revenue or reader revenue.
The big problem here is that ad revenue and reader revenue strategies have been competing with each other.
When we first launched our websites decades ago we relied only on ad revenue. And we became good at getting traffic to our websites which meant we could increase ad revenue.
Still, we didn’t know the readers, the users, they were only “traffic” to us.
During the autumn of 2014 Mittmedia launched its first paywall.
This was before I started to work for Mittmedia. But I’m told there was one — big — problem: We didn’t want to do it.
The whole reason for us to launch a paywall was to protect the print revenue. We had raised subscription prices and were afraid print customers would cancel their subscriptions and consider our online content sufficient.
The goals were pretty simple:
- 50 % of published articles were to be put behind paywall.
- And we wanted 50 % of print subscribers to activate a digital account.
We succeeded at both goals. The first goal — to put half of our news content behind a paywall wasn’t that hard to do.
But. And this is a pretty big but:
The articles behind paywall was our least read articles by our paying customers.
So the content that we put behind the paywall was our worst content. Content no one wanted to read. Not even the customers that had access to it. They preferred our free content. Content that was made with data-driven focus to get page views. The content behind the paywall was the content that originated from the printed newspaper.
So we had introduced a paywall with holes so big that no-one noticed it. And we weren’t getting any digital reader revenue.
When I started to work for Mittmedia — one of my first decisions were to scrap the 50 percent rule and effectively scrap the paywall and start experimenting with reader engagement.
Between 2015 and 2016 we grew our digital audience to record numbers. Loyal users that gave us ad revenue. But none of them paid for access.
And we weren’t really getting paid for our traffic increase. Instead, the ad revenue started to go down since other players could gain much more traffic than we could. In a classical supply and demand situation, the price on ad inventory went down.
In 2016 we launched our paywall again. But this time we saw digital reader revenue as the main strategy. The market was ready for digital subscriptions and we, as a news organization, we were ready for that as well.
And we built our paywall with a data-driven focus of increasing digital reader revenue instead of keeping print revenue as we did last time — when we failed. This time failure was not an option.
Since 2016 Mittmedia have managed to build a digital subscription business and been able to increase the ad revenue at the same time. Without increasing traffic.
I’m going to tell you how we did it.
Let’s start by zooming in on the traffic. That was our first problem. That we saw it as traffic and didn’t care who was behind the traffic.
It’s kind of ironic that this isn’t something that we would do when it comes to the printed newspaper. There, our relationship with our paying customers was our number 1 priority.
So, in order to build our business on loyalty, we needed to focus on the users and see user data as our number 1 priority.
We need good data to:
- Produce and distribute relevant content. Content that the readers want and make them come back for more.
- And monetize on the users by selling digital subscriptions and building target groups in order to sell ads.
And since data is crucial for our business we built our own data platform.
Our own data platform — for better products and insights
This is Mittmedia’s digital ecosystem — a system that our in-house development team have built and are continuing to develop every day:
In the middle we have the data platform. We call it Soldr. Soldr is really our heart and brain at the same time. It’s where we store all our data.
If we want to extract value from interactions, content and users we need to own our data.
The opposite of building your own data platform is the black box solution — like Google Analytics. A system where Google tells you what data is important.
The most important component of Mittmedia’s data strategy is the demographic and behavioral modeling of our user base. To understand the data.
To stay innovative we must understand what values the users find in our products, and why they chose to engage with us in the ways they do.
That’s what Soldr enables.
In Mittmedia’s digital ecosystem there are other components like
Aracua — our content creation and curation tool. This is the interface where our journalists spend most of their time.
Or Reacher — the system our ad sales staff use every day to sell ads and a system that finds the right target audience. Reacher is a cornerstone in our advertising strategy.
This is a good example of an application made possible with the help of Soldr:
One of our most important KPI’s is daily user activation. Both our subscription business and ad business is dependent on frequent use of our products. If people use our products every day the risk of churn is heavily reduced.
This simple — yet powerful — tool that we call “The User Activity Tracking Map” visualizes that to all our newsrooms. It shows our paying customers as dots on a map.
We also plot the coordinates of the content being published. That way it’s easy to see the correlation between the users and the content.
When the day begins all the customer dots are red. As the day goes along and the customers start using our different products the dots turn green.
Your goal as an editor is to make sure that our customers find our content to be so relevant that they use our products every day. When day turns into night you want the map to be filled by green dots.
With the help of this tool we have seen daily activation rates go up from around 50–60 percent to 60–70 percent.
DATA INFORMED CONTENT PRODUCTION: Data insights to produces the right content
Soldr helps us produce and distribute relevant content. Content that makes our loyal users stay loyal.
As you can recall we failed at our first digital subscription attempt? We learned many things from that process. Perhaps the most important one was this.
Print and digital are two different offers with different needs. The content that works in print does, in most cases, not work online.
But we know what works, thanks to our data platform Soldr. We extract data from Soldr to answer the questions:
- What content make users convert into digital subscribers and
- What content make them stay active in our products.
This table tells us that:
But I don’t expect you to be able to read it since it’s too small — and also it’s in Swedish (and it’s a bit dated). But even if you were able to read it it would be hard for you to take action on it since it’s not that easy to spot important patterns.
Bubble charts are much easier to use. Here we see the efficiency of our content as a bubble chart.
The bigger the bubble the more efficient it is to convert loyal users into paying users. The y-axis shows the number of articles and the x-axis shows the number of logged in users. We can see that local business news has big potential. Yet we didn’t produce that many local business articles.
So a conclusion we can draw from this chart is: If we produce more content on local business — more users will convert into digital subscribers.
But to make 600 journalists read bubble charts and take action wasn’t something that we thought was possible.
Inspired by Norwegian media company Amedia we translated the charts into Content Guidelines.
Our Content Guidelines highlights what areas we are to focus on and recommendations in every area. In this example, we took local business.
What local business content are we to produce more of? What content we should stop producing? And other success factors.
To be this concrete when working with the news room has been one of the most important initiatives.
We now have a group of professionals that we call the Content Development Team. They work full time with our content.
They interpret the data and help our local journalists — like editors and reporters — to take action on the data.
The team members are in contact with our local newsrooms on a daily basis. Therefore they can also find best practices and spread that in the entire news operation.
Let’s go back to the bubble chart again. What kind of content makes most users log in? It’s house sales — real estate. How could we start making more house sales articles?
We could, of course, tell the newsrooms to start making more articles on house sales. But we could also take a more unorthodox approach — but much more effective…
Automated content helps our journalists focus on the right stuff
We built a robot that produces articles about every house sale in our regions. This is how it looks like on one of our news sites (it’s in Swedish language originally so this is just a translation):
It’s automatically published whenever there’s new data on houses that have been sold. It looks like any other story published in our local news products. The big difference here is that no human being has been involved in producing or publishing the article.
I know that some of you are probably surprised that information like names of sellers and buyers and the price of the house are open to the public. But Sweden is a very transparent society. And we have for decades published lists like this in our newspapers — with the same information that we built the Homeowners Bot on.
But it’s a very different skill to publish lists in the printed newspaper once a month compared to publish automatic articles every day online.
A couple of years ago we founded a company, United Robots, with two other partners, to develop a platform for automated content. So in this case, with the Homeowners bot, we paired Mittmedia’s content developers and United Robots system developers. Their mission: Let the data tell a story.
And this is how the bot that they developed, work:
We get the data to the Homeowners Bot from Sweden’s Land Registration Authority. They include data points like the name of the seller, the name of the buyer, the address and the price.
When the data is cleaned our algorithm analyzes the data to find the best angle. Is the house unusually expensive? Is it a big house? And so on.
It works kind of like the normal journalistic process when a human works on a story. You collect the information, you decide on the best angle and then you write your story. Except, in this case, every step is automated.
We are very transparent about the origin of the article. The byline on all automated content in Mittmedia reads “Mittmedia’s Text Robot”.
How do you think robot journalism is performing? “Mittmedia’s Text Robot” is our most efficient “employee” — if you would compare it to a human beeing.
It produced 64 000 articles and got 3 million logged in page-views between September 2017 and September 2018. That’s impressive by our standards.
We now have quite a lot of automated content: Sports, local business, traffic, public transportation, weather alerts and so on.
It’s content that we saw a clear demand for in our data. And we are continuing to develop new automations.
DATA DRIVEN CONTENT DISTRIBUTION: More relevant news with the help of personalization
By producing great content that makes the user convert to paying users and content that makes them stay, we have built a nice foundation for our subscription and advertising business.
But producing content is only part of the equation. We also need to have an efficient distribution. I’m going to show you how Mittmedia builds personalization in our data platforms and products.
Why have we chosen a strategy to use machines and AI to personalize products? It’s about relevancy. And relevancy is deeply personal.
The digital editorial products that we see today are, with very few exceptions, built upon the same principles as the printed newspapers.
Everyone gets everything. It’s the same selection for everyone. A selection made by an editor.
We haven’t taken advantage of the possibilities of data. Until now. With the help of data, we can improve the content relevancy for our logged in users.
In November 2017, we decided to begin experimenting with personalization. Our aim was to enhance the relevancy of the selection rather than building a product that is 100 percent personalized.
We still believe that a journalistic product must give the user a big variety of content. The goal we set was quite humble:
Today a customer enters a Mittmedia product at any given time, the customer will look through a list of articles and find for example one article of relevance per visit.
Mittmedia needs to create products that will give the customer the opportunity to at any given time, find and consume two articles of relevance.
This is not rocket science.
So, how did we do it? On a conceptual level, this is how Mittmedia is approaching personalization:
By using data from logged in users we are able to create different user profiles that the system can learn from. These profiles are translated to cluster data, and each user in the system is assigned to at least one cluster.
A user belonging to a sports cluster will see more sports content in Mittmedia’s products.
When we meet new users, or when a user behavior changes, the algorithm re-assign a cluster belonging.
When we began the experiment we only had 9 clusters. Today, in December 2018, we have over 560 user cluster combinations and we will see more clusters being defined as we get to know the users even better.
I believe the cluster approach is a controlled and responsible way of applying personalization in a journalistic product. But going forward I also believe that we will see individualized personalization.
In the process we also realized that our existing homepage structure wasn’t optimal for personalized content. We ended up redesigning the whole page and gave it a more feed based layout.
In September Mittmedia launched the machine driven and completely autonomous personalization of our news sites.
So today we have only three article teasers on our homepages that are manually handled, all other teasers, and there are 100’s of them, are automatically displayed depending on what cluster you belong to.
And we can now conclude that the machines are doing a better job in this aspect than humans.
How about the results?
We see a significant increase in the number of clicks (CTR) from our start pages, from 1.6 to 2.0. I expect this number to continue to increase as we continue to iterate on the personalization algoritm.
In fact: All the metrics that we use to evaluate the personalization of our start pages shows positive effects.
- We have more daily active users.
- The users spend more time on our products.
- And the overall number of page-views, generated from our paying users, have increased.
I believe this is a great example of the use of AI and machine learning in media products.
To quote Ryan McCabe, data scientist at Spotify:
Machine learning has moved past the high-school sex phase (everyone talking about it, but no one doing it), to much broader implementations across industries including media.
How we monetize on loyalty
We have now seen how content drives loyalty. Loyalty that we can monetize in different ways.
It wasn’t that many years ago when we celebrated traffic records. But we didn’t care who were behind those page-views. We didn’t need to know them to generate page views. We only wanted their clicks.
But to generate revenue today we must know the users. And make loyalty your goal.
We have a strong and loyal audience today. Almost 90 percent of Mittmedia’s page views are direct traffic.
That’s a pretty unique position when you look at it from a global perspective.
In Reuters Institute Digital News Report 2018, that was published in June, they looked at how readers find their news. The Nordic region really stands out. Finland, Norway, Sweden and Denmark have the greatest proportion of direct traffic in the world.
In the report they come to this conclusion:
These differences in preferred access points are critical. They show that Nordic publishers still have direct relationships with their readers, making it much easier to charge for content online.
It’s all about relationship and loyalty. We have, indeed, a strong tradition for reading and subscribing to newspapers.
In Norway and Sweden around 30 percent of the users pay for online news. And the numbers are still growing. We stand out if you look at the other European countries. Only 8 percent have paid for news in Germany for example.
Mittmedia’s subscription offers
This is how Mittmedia sell our news products today. We have 3 main offers.
First offer we call “Plus”. It’s all content behind paywall. And today all our journalistic content is behind paywall.
The price is 99 Swedish krona — or roughly 10 euros a month. This has become the standard pricing for entry-level digital offers in Sweden. Spotify and Netflix are two fine examples of this.
Then we have the “E-paper” offer that includes both e-paper and behind paywall material. The most expensive by far is the Print-offer that includes everything.
Many publishers have included the e-paper in their basic offer but our data shows that this is not relevant for the target group.
Let me explain why: Have a look at this:
This is the age distribution for e-paper and print subscribers. They have roughly the same age distribution. The average print customer is 70 years old and the average e-paper subscriber is 61 years old.
The Plus subscriber — that only pay for paywall access — is 20 years younger. That’s why we don’t include the e-paper in the basic offer.
Our acquisition strategy is pretty simple.
- We optimize our websites and apps to attract new Plus-customers.
- And we try to transform print subscribers into e-paper subscribers.
What we are working on now is a way of getting more revenue from the Plus-customers.
We are looking into adding a Plus Premium offer to be able to move the customer up the value chain. In the Plus Premium offer we could offer features like log-ins for the entire family and content like live streamed sports games.
This is perhaps our most important chart:
It shows our logged in users per week. As I mentioned previously in this post, with the activation map of green and red dots, we focus on our customers using our digital products. The reason is simple: People who use our products pay for our products.
Half of our weekly digital users are actually customers who subscribe to our Print offer. It’s perhaps a bit surprising. But they also are heavy users of our digital products. That’s important. Especially when having a transformation strategy. We need to get as many of our print customers aboard our digital channels in order to keep them in the end.
The growth is of course strong also when looking at the digital subscription revenue. 81 percent increase in 2016, 100 percent increase in 2017 and January to September the growth was 48 percent.
Our main revenue is, as is the case for almost everyone in our industry, the printed newspaper. But we feel comfortable entering a digital-only business when we get to that point. And we are rapidly nearing that point.
In 2020 half of Mittmedia’s profit will come from digital products. Even though the print revenue will be double that of the digital revenue. The print costs will simply eat up the print revenue making the profit margins much better on the digital products.
Our digital future success is of course totally dependent on us delivering growth every week.
We convert almost 2000 users into Plus subscribers every week. And the trend is increasing. So conversion is not a problem.
But net gain is. Too many customers today are canceling their subscriptions. We need to get churn under control in order to continue our steady growth.
That’s why our focus now is retention.
And our strategy to handle retention is loyalty.
Mittmedia’s 5 digital transformation insights
- Focus on the user. Encourage all employees to always take the user’s perspective.
- Try, fail and learn. Repeat. It’s okay to take risks, fail fast and learn. Actually, it’s necessary in order to go forward. More action and less planning!
- Encourage real innovation. We are in the middle of a transformation journey. Old truths don’t apply. We must find bold solutions and be innovative. (And fund R&D accordingly)
- Strategy and communication go hand in hand. It’s not enough to have a clear strategy. The strategy must be communicated. Over and over again.
- Know your mission, your purpose. We believe in the power of journalism. We want to make our local communities a better place for all.
Robin Govik is Chief Digital Officer at Mittmedia. You can reach him at firstname.lastname@example.org