Have you ever thought of canceling your subscription because you only have time to read one story a day at best? A digital team of Aftenposten, an over one hundred-year-old newspaper from Norway, has found a solution. They are developing a synthetic voice to voiceover all of their stories, which will allow subscribers to also listen to content while shopping, exercising, driving, etc. This greatly increases the value of the subscription for their paying readers.
And that is just one of a few changes the team is introducing in an effort to make 250 thousand of their subscribers happy. Aftenposten is a Norwegian daily, part of Schibsted media house, founded in 1860 and the biggest print publication in the country. However, the team’s focus now is on digital development. The paper launched a digital subscription in 2010 and is planning to reach 150 thousand digital-only subscribers this year. In September 2021 they were just a bit above 130 thousand.
According to Karl Oskar Teien, Director of Product at Aftenposten, the transformation includes work across three main shifts that happened in the world of media in the last few decades.
The shift from physical to digital distribution is the first one the team has decided to tackle. Aftenposten’s product team believes this is not just about the change of distribution means, but global competition for screen time. Every other product nowadays literally feeds readers the information. All, but the front pages of the news media. Aftenposten’s team found out that half of the front page visits end without a single article view. From there, the solution was to turn the front page into a product.
The shift from content scarcity to content abundance led the audiences to seek out having rather than getting informed. And this is another change the team of a Norwegian daily has accepted and has already started changing their editorial approach.
Last but not the least, the shift from one single method of distribution to several. Staying on a single platform will soon (if not already) become disappointing for the paying members and subscribers. Reading is not enough anymore. And that’s where the synthetic voiceover solution comes in.
The digital transformation led by Teien is still ongoing and the changes are being implemented as we speak. Teien joined the paper a year ago to adjust the media’s digital product to the needs of subscribers.
The Fix met Tein at the IPI World Congress in Vienna in September 2021 and spoke to him about the overall approach to digital transformation, tests and forecasts for the new features and managing a digitally focused newsroom.
This interview has been edited and condensed for clarity.
The Fix: What is the story of your digital transformation? At what point did you realize that you need to change in order to provide the relevant service for your readers and subscribers?
Karl Oskar Teien: We started digging into these things when I joined Aftenposten, around a year and a half ago.
The first phase was just about identifying the problems we are really trying to solve here. I should say that Aftenposten had already done a positioning analysis. Over the course of the fall of 2019, we ran a big project on figuring out what was a profitable position for us. What we mean by that is what type of journalism profile we need to have in order to succeed with growing more in terms of subscribers. In other words, what are the drivers of people’s willingness to pay? At the end of the project, we had a clear idea of what our value proposition was. And the value proposition that we decided on was to be an explanatory news experience. That means that we are not just providing the news within certain categories but we are focused on explaining and giving new insight, the new perspective.
The Fix: What made you think you needed to change your editorial approach? And how did that change lead you to the change in the product?
Karl Oskar Teien: Sometimes we talk way above our readers’ heads, and it’s hard for them to understand exactly what we mean. We realized that if we want to grow more, we need to reposition ourselves.
So it was kind of at the tail end of that repositioning that I said, well, we have to do something about the front page. That’s when we started first looking at the data and looking at how people use the front page today. What user groups do we have? How do they interact with the product? How often do they come and visit us? Based on that we did a discovery process. We started to figure out ways that we can serve them better. Now we’re in the implementation phase. We have built a lot of the infrastructure for it and now we are just establishing new principles for how the front editors work. We are introducing a few new concepts.
The Fix: You’re in the process of launching a synthetic voiceover for the key stories to create another option to consume your content for your readers?
Karl Oskar Teien: It should be ready in about three weeks. We have taken a lot of sentences from 2000 articles that we’ve written over the past few years. We are selecting articles with broad differences in pronunciation. We’ve got eight hours or something of speech data and fed it into an AI algorithm that will then generate the voice that sounds like our main podcast anchor. She was, of course, willing that to happen as well. She thought it was an honor that she would be the voice of Aftenposten given that we use it for the right purposes. I think it unlocks some other opportunities later on. As I mentioned, I think there’s an opportunity to extend the reach of our text content.
The Fix: What was the testing process for this new product?
Karl Oskar Teien: We have done user testing, we tested different scenarios. We got feedback on how easy it was to understand and how it sounded (human voice – The Fix) compared to a very robotic voice. And we got qualitative feedback. We’ve also run a series of AB tests with a few thousand users that get exposed to an article and they either get a robotic sounding voice and the human voice. We looked at how different the engagement level is. How much longer do they listen when it’s a human versus when it’s a robot? It’s interesting to get a baseline. Then we’re going to look into how we can close that gap between the engagement with a human voice and synthetic voice, when we put in all the framing. Who knows, I think we will probably never reach a point where it’s as good as the human voice. But if it’s 95% there, it opens up some opportunities that we otherwise wouldn’t have. I think it has value to experiment with it and be at the forefront of a trend that we know is important. Not necessarily that we will immediately see a lot of subscription growth, direct, faster results of it. But in a few years from now, it will probably be very hard to tell the two apart. That’s when we should be ready with a user experience around that.
The Fix: How would the product look and function? You said there would be a play button in the article? Would it be possible to create a playlist on the website or download it on iTunes?
Karl Oskar Teien: It’s so new that we don’t really know. But our bet is that we need to be visible everywhere. If all of a sudden the audio appears in some articles, but not all of them it can cause frustration. We want to give our readers the ability to play every article. Then we should experiment with doing that right from the front page on the teaser itself.
But as I mentioned, I think the real value comes when you start thinking about solving a problem that people identify with a lot. Let’s say you are paying for a subscription for 30 euros a month and you realize that you’re only reading two or three articles a day. It’s a good reason to cancel your subscription, right? But if you realize that, okay, I’m reading two or three articles a day, and then I’m listening to five while I’m commuting, all of a sudden the value of the subscription is higher.
We’re extending the use cases for the text that we produce. It feels unfortunate that we produce so much great journalism every day that very few subscribers end up reading. So how can we make sure that we help people make use of what they’re paying for essentially.
The Fix: Are there any other monetization opportunities to come with this product?
Karl Oskar Teien: The Washington Post does this already. They have advertising at the beginning of synthetic audio clips. If I want to listen to an article, first I’ll get a programmatic ad in the beginning, and then one at the end as well. So that’s one way of monetizing it. I doubt that we will go that route, I think we will want this to be a subscriber-only feature, and something that we can promote, not just as a kind of part of the sales pitch, but something that will help retain users. So it’s really gonna hurt when you cancel our subscription, because all of a sudden, you can’t listen to your favorite articles from your favorite podcast voice anymore.
The Fix: What other changes are you working on right now?
Karl Oskar Teien: We are changing the way we structure our content on the front page. We are now bundling together stories in a much clearer way. So instead of having just a lot of different articles, we’re putting them together in groups, and we’re helping users navigate within the topic rather than navigating within a section. Because division by sections is an outdated way of thinking about content navigation. It’s based on how the newsroom is organized and where people work. We said, well, can we instead help users navigate by topic based on what the story is?
(Before that Aftenposten had already launched a mobile version of the feed-based frontpage – The Fix)
The Fix: Mainpage redesign sounds like a big deal. What was the reaction of your subscribers to this change? Did they appreciate it or was there some sort of denial?
Karl Oskar Teien: It’s a little bit early to say. We launched a redesigned version on mobile just a little less than a month ago and are introducing some new features right before the election (Parliamentary elections took place in Norway on Sept. 13, 2021 – The Fix).
A lot of what I presented here were results from experiments. The experiments show us that the users who typically are lower frequency, are big fans of the new concepts. The higher frequency users, who are more used to things being the way they are, super users, are not as positive. They tend to have an approval rating of 75%. As for the lower frequency users, their approval rating reaches from 85 to 90%.
The Fix: You said that we often act on what we, as journalists, think is important, and not what is important for users. How do you decide what is important for the users? Do you seek out feedback from them on your editorial choices and product changes?
Karl Oskar Teien: That’s a good question. I would say this is still mostly an editorial decision. Our editorial choices are less affected by what’s trending and what’s most read.
We think one of the main reasons for subscribing is a need for someone to tell you what you need to read, rather than what others are reading. In print you would have four or five articles highlighted. That’s kind of what we do online as well. But we also try to summarize the stories briefly and do it more the way that you would consume Instagram stories or something like that.
The Fix: You said that you mostly target the most frequent users with these changes, people who see your website at least 3 times a day. Have you probably looked into the AI solutions to customize different types of features for different user categories?
Karl Oskar Teien: We have some of that today. Our newsfeed is editorially controlled, but it’s ranked algorithmically. What that means is the first three to five stories, the top stories, are manually selected. But the stories below are lined up automatically by a queuing algorithm taking in the editorial signals. The editors assign a certain amount of points to each story based on its news value, the timing of its publications, articles’ performance on the website compared to where it’s placed on the front page. For example, if the story performs very well, for being relatively low on the page, it will be bumped up higher.
So we can say that an article is a news value 1, or it’s a news value of 5. If its news value is 5, it means, from an editorial perspective, it is really important. It will always be ranked higher and it would be the popular news. So that’s how we combine editorial signals with some kind of signals that will help you find articles that others are reading. We think that balance is important here, as we have a responsibility for telling users both – what is important from an editorial perspective, but also what other people find important.
The Fix: I would assume that the correlation between the story performance and the editorial value you give the story also provides you with lots of data. How do you use this data? What editorial decisions do you make based on this information?
Karl Oskar Teien: So we tried to focus less on the clicks, because that a lot of times depends on how good you are teasing the story, how good is the title and where it is on the homepage. Instead, we try to look at where people drop off in the article. We want users to follow throughout the whole story, and then drop off more toward the end of the article. But if we see that the drop-off is really high, then that’s a clear signal that either the story isn’t interesting in the first place, or we didn’t tell the story in the right way. We’re seeing a lot of difference, for example, when we experiment with storytelling and have a more visual approach, with text or images unfolding. We’re actually exploring ways of doing that much more than we are today as a way of getting users through the whole article. Apart from that, we’re quite focused on seeing what converts people into becoming subscribers.
The Fix: You also mentioned that you use some of these indicators, including sales, as KPIs for the newsroom. There is a heated discussion about this in the media sector (There are media managers that believe sales KPIs to be inappropriate for the editorial staff – The Fix) How does that work? Can you elaborate on that?
Karl Oskar Teien: We evaluate different things. A part of that process is to evaluate the number of sales and the extent to which people read through the whole article and, of course, total pageviews. The sales part of it is, I think, something that we don’t want to separate from the rest of the editorial. You might say, well, that’s kind of problematic, because then you have a very commercial mindset in the newsroom, when we should be covering stories, regardless of how well they sell.
The general evaluation of our journal entails a combination of factors. We want a high number of journalistic pieces that really set the agenda for public debates. And we mentioned that as well. You can’t measure to what extent it set the agenda, but with some stories, you just know, this became the thing that was debated on TV, uncovering some big story that otherwise wouldn’t be covered.
We measure success in a lot of different ways. I don’t think there’s one specific answer to that. I think it is important that the newsroom is aware of the fact that we are here for our readers. That’s a change of mindset that has happened with us thinking more about subscriptions, sales and serving our customers.