How do media companies stay competitive?

We Like Mags and SPRYLAB in conversation with Olaf Deininger, journalist, digital expert and publisher

What skills do media professionals such as journalists and editors need to have in order to turn readers into regular users and, ideally, to sign up for paid subscriptions?

Thoroughly researched stories and good texts are still important, but today data plays an important role in attracting customers, as do the appropriate tools. Tools that provide important information on how to identify target groups and their interests in order to make them attractive offers.

Money follows Data?

In We Like Mags and Sprylab's second special on audience development, hosts Christian Kallenberg and Benjamin Kolb talk with journalist, digital expert and entrepreneur Olaf Deininger.

Among other things, Deininger talks about the fact that users are becoming increasingly volatile, while at the same time media behavior is becoming more volatile and competition among media companies is becoming tougher. That's why AI-based solutions are now necessary in everyday editorial work if newspaper publishers don't want to be left behind.  

Related links

What awaits you in this episode

  • 03:46 - What can journalism-as-a-service do?
  • 05:32 - Impact of technologies on journalism
  • 08:44 - Don't be afraid of artificial intelligence
  • 10:53 - When artificial intelligence writes stories
  • 15:00 - Artificial Intelligence for Audience Development
  • 19:47 - Filter bubbles
  • 24:49 - Data-driven journalism
  • 26:45 - Like looking into a crystal ball, only better: predictive analytics
  • 30:20 - Community Building: Providing a Digital Home for Readers
  • 37:39 - Nothing going on without a community manager?
  • 39:42 - How does AI fit into communities?
  • 40:42 - Olaf Deininger private


Christian Kallenberg [00:00:10]:

Welcome, dear listeners, to the new episode of the Audience Development Deep Dive by SPRYLAB Technologies and We Like Mags. One of my first guests on the We Like Mags podcast was Benjamin Kolb, CEO of SPRYLAB Technologies, with whom I talked about audience development and other topics, and who is sitting across from me now. At the end of our conversation, Benni and I had the feeling that not everything had been said about the topic, and that's why we decided to dedicate a series special to the topic of audience development, in which we would interview guests together about the various aspects of this increasingly important topic. And, Benni, I'm really, really grateful that you, as a digital expert for publishers, are co-hosting this format with me. So tell me: What are we talking about today?

Benjamin Kolb [00:00:59]:

Yes, hello, Christian. Our guest today is Olaf Deininger. Olaf Deininger is a journalist, author and digital expert. And with Olaf Deininger, we're talking about audience development, of course, but also in particular about artificial intelligence methods in this area and, in particular, about communities.  

Christian Kallenberg [00:01:21]:

Here we go. Hello, Olaf.  

Benjamin Kolb [00:01:24]:

Hello, Olaf. Pleased to meet you.  

Olaf Deininger [00:01:26]:

Yes, hello everyone.

Christian Kallenberg [00:01:28]:

Hello, thank you very much for taking the time today. And, Olaf, we really do have an enormous number of topics in mind, so let's start off sportingly with the big sweeping blow: How do you think journalism has to change in order to remain contemporary and endure in the long term?

Olaf Deininger [00:01:44]:

Wow, that is of course a huge question - the ideal question to start with. You could write books about it. There are probably already quite a few on the subject.

Christian Kallenberg [00:01:55]:

But not from you.  

Olaf Deininger [00:01:58]:

Exactly. That's a question that many media companies - most of them, perhaps - are probably asking themselves. I think it's safe to say that technology and other factors, different user behavior among users and readers, and so on, are making this question much, much more relevant than it was five or ten years ago. We are experiencing a technological transition, and the first thing that comes to my mind is, of course, that we have to think about user values and a stronger target group orientation, and in this context, of course, also about new formats that perhaps satisfy user expectations that have not been completely satisfied to date.  

Christian Kallenberg [00:02:41]:

What formats could they be?

Olaf Deininger [00:02:42]:

We see a media industry that is very much focused on news. That is also true. That is also the core task of all media, basically and always. On the other hand, we also see that there are certain, relatively many information needs that cannot be covered by news at all. Let me give you a small example: Where I live, there might be a big annual folk festival and I'd like to go to it. And then at some point the question arises: "Where can I actually park this year and what does the parking ticket for the afternoon cost?" Then it's usually something that I might find in some news item in some daily newspaper. It is then hidden on line 230 and I don't want to read through the whole text to find out exactly this information. And from my point of view, this raises the question of whether this kind of information can be covered by a news item, a report, so to speak, or whether new formats are needed to give readers, users, precisely the information they need on this Saturday afternoon, namely: Where do I park? Or: What do I have to spend for it?  

Benjamin Kolb [00:03:46]:

At the end of last year, you wrote an article about journalism as a service. I found it very exciting. At SPRYLAB, we know a lot about software-as-a-service and platform-as-a-service and infrastructure-as-a-service, but what is journalism-as-a-service? What is that for you?

Olaf Deininger [00:04:04]:

First of all, we see that journalistic activities are becoming more and more differentiated. If you look at modern newsrooms, there are, so to speak, many, many different activities, different types of editors. In the past, there was the writing editor, the page-turning editor. Today, it's not so simple anymore. There are many different functions. "Search engine optimization" would be a keyword there, "news research," "new format development," and so on. That's what we're seeing on the one hand, and on the other hand, in the supplier market for freelance authors and platforms, we're seeing suppliers who basically offer complete solutions. We see the keyword "Creator Economy". We see platforms that cover certain functionalities for me as a journalist, so that today, for example, I can configure a newsletter within two hours and do this via a platform that, in principle, helps me to charge a membership fee for it, for example, or a subscription fee, and then collects it directly from the recipient accordingly. I don't have to worry about billing, invoicing and so on, but can start immediately with the content of my newsletter and everything else that goes with it, monetization, invoicing and so on, is done for me by the platform. That would be a solution for journalism-as-a-service, for example.  

Benjamin Kolb [00:05:32]:

From your point of view, what is the consequence of technology in connection with journalism? What impact does technology have on journalism, on editorial work, and above all on the end product for the consumer?

Olaf Deininger [00:05:47]:

So I'm going to say quite fundamentally: There is no area of the editorial process chain that would not be affected by this in some way. Technology allows - let's start with the front end - new editorial, content formats that were previously inconceivable, for example, that I can integrate some data on maps and thereby make it usable on a completely different level. It supports editorial processes by automating or semi-automating certain activities. A good example is an application that automatically and constantly keeps an eye on the weather forecast of a local newsroom and informs the editors in advance, for example, when a storm is brewing, and sensitizes the newsroom to the fact that there may or very likely will soon have to be coverage of how the fire department is going to pump out flooded basements and things like that. In other words, we can go as far as research - that would be another area, perhaps, that AI solutions systematically provide me with preliminary research, create initial rough texts, and relieve me as an editor or as a writer at that point when it comes to taking initial research steps. And I think I could expand the list as much as I want now. Perhaps the three points are important, so to speak. There is the possibility of developing new formats that are relevant to the target group, there is strong procedural support, and thirdly, there is support in the area of research.

Benjamin Kolb [00:07:28]:

And these are the points where we, I think, also have a bit of a common vision that technology can actually remove the hurdles for journalists or editors and thus give them back a certain freedom that was actually first created by technology, where we basically have the phenomenon that editors now have to deal with search engine optimization and other digital hurdles in their work chain. Do you see it the same way, that technology can bring journalism back to the actual work, i.e. creative work?

Olaf Deininger [00:07:58]:

Absolutely. I know companies where highly qualified editors are busy rewriting press releases in the specialist area, putting them in the subjunctive and shortening them for print sections, some of which are 20 or 30 pages long. For me, that's one of those activities where I don't really make the best use of my editors' skills. These are activities that I can now map relatively well, at least in part, using technology. And when I do that, I create resources so that my highly qualified editors are occupied with what they are actually best at, namely conducting interviews, doing background research, evaluating things. Anything other than cutting down press releases.  

Christian Kallenberg [00:08:44]:

Olaf, you just mentioned that AI can be used to roughly pre-formulate texts. Of course, this is always a bit of a bugbear for many veteran journalists. What is the reasoning there? Why shouldn't we be afraid of AI?

Olaf Deininger [00:09:00]:

So basically at one point I would disagree. Technology inherently is always a dualism. Of course, every technology is also some kind of streamlining technology or contains the possibility of streamlining things. So I saw in the 90s how the whole prepress disappeared. I can still remember large print shops, halls the size of soccer pitches, where there were lots of light tables, where printing templates were assembled, at Tusch-Druck in Vienna, for example. None of that exists anymore. Technology has ensured that not only has an industry changed completely, but that an entire area, namely prepress, has disappeared. That is one side of technology, which is always there and cannot be avoided. But the other side of technology is that it can actually make things more productive, that it can ensure that people, as I have just described using the example of press releases, are relieved of activities that require little qualification. In this respect, this dualism must be properly controlled. In this respect, I would give my colleagues the message that, as is always the case with technology, you also have to be prepared to learn something new. You have to penetrate, understand and master this technology in order to be able to use it optimally in your own interests. And at the end of the day, bottom line, I believe that at the moment, in the phase where we are in the media, the opportunities that arise through AI for the media and for journalists are actually much greater than the dangers. I could well imagine - and I'll come back to the beginning very briefly - that editors who now deal with press releases would be very pleased if an algorithm took a large part of this work off their hands, so that they could take care of really journalistic things. In this respect, I don't think there is such a huge risk, but rather more opportunities.

Benjamin Kolb [00:10:53]:

Let's stay with the topic of texts and AI or text generation and AI. I know many people in the field of data-generated texts, i.e., for sports news, sports tickers, stock market tickers and the like, where you can create very interesting texts from a lot of data and certain text modules that a reader can't necessarily distinguish from a text written by an editor. But now the question arises: Is this already going further nowadays? So do you have a bit of an idea of where we are at the moment? Can real prose texts already be generated? Your opinion on the subject.

Olaf Deininger [00:11:29]:

So where AI, I think, can be very useful, I use that myself from time to time, is in the area of preliminary research, which I already said. There is a nice tool called "AI Writer". I think it's from the university in Nuremberg. I would have to look it up again. You can enter three or four keywords and it does a basic search. And my experience is that if you use it, not always, but in two-thirds of the cases you get a very good impression and maybe you've saved yourself a quarter of an hour's time. This is a very practical and functional and good application. It's relatively inexpensive, and it makes sense to run it as a first step if you're new to a topic. There is a second tool, developed by Schickler, the management consultancy, that can shorten texts very well. You can use a slider to enter the percentage to be shortened. They have made certain discoveries that, for example, the first sentence in a message is actually extremely important, which means you can set a small checkbox so that the first sentence is excluded from the shortening in any case, and things like that. I also think it can be extremely practical to work with such a thing. The fact that the editor then has to go over it again and perhaps take another look is a bit in the nature of things, but he has then already done a work step at least partially. As far as text generation itself is concerned, this is still a field that is relatively in its infancy. Now I come to the example of Daniel Kehlmann. Daniel Kehlmann is a relatively well-known author in Germany in the field of fiction. His most famous or his best-known work is "Measuring the World," where he describes the life of Alexander von Humboldt, among other things. And he was able to experiment with an AI called "Control", and that was supposed to be a bit of a ping-pong game, that he writes the first sentence, creates, the AI then does the second, he does the third, AI does the fourth and so on. And he experimented with that for a while like that. Practically, the context was to create short stories, and his conclusion is that after one page of A4 text at the latest, it somehow becomes a bit grotesque and then often fails, because the AI loses the context so to speak. This is a good example to show where technology currently stands in the field of literary text production. But this is all technology that is still in its infancy, and I believe that we will see enormous progress in the next one, two, three years, especially in things that can be structured well, such as stock market results, stock market developments, but also the results of soccer matches or sports in general. It makes sense for me to automate things - match reports, reviews or texts on the subject of match results - and perhaps in this way get into a position where I can present local sporting events in the media at a cost that is feasible, whereas if I had to pay authors or editors for this, I would no longer be able to finance it as a publisher. So, from my point of view, this also creates opportunities to create and deliver content that would no longer be economically viable if we were to do it with our hand on our arm, so to speak.  

Benjamin Kolb [00:15:00]:

To leave the area of text generation for a bit: We have divided up the possibilities of the areas for artificial intelligence for our customers or for our products along the value chain of publishing, i.e. from topic identification, topic discovery, what can you do with news or seasonal topics, via what we have just talked about, i.e. the support of the actual journalist or the editor, be it SE optimization or text generation, to what we call "performance enhancement", which are actually the topics of audience development. So how can I bring more target audience into my business model with my content? How can I gain more subscribers? How can I also do churn prevention? What do you see as the application areas of AI that you see as potentially promising for journalists in the coming years?

Olaf Deininger [00:15:59]:

This is also a very broad area from my point of view. It might start with the fact that AI gives us the opportunity to come up with a completely new way of finding topics, generating topics, because we have new ways of evaluating interest patterns and in this way possibly discovering patterns in terms of interest in content that we would otherwise not discover with the classic methods. That would be such a big area that I'm sure we'll see. A second area you already said. Basically all the things where it's about tagging content, images and so on. Can I consider automating those to a large extent. Image recognition is a relatively well-developed area, so to speak. Some people say that recognition is already a commodity, a standard product. Here, too, we could think more about the automatic generation of metadata and keywording. Another example of what you said: Search engine optimization is, in my view, also largely a rule-based discipline. And whenever it is rule-based, I have at least the option of possibly automating it, and I can also relieve my colleagues by automatically optimizing the content of web pages, at least as a suggestion for the editor, either partially or completely for search engines. The editor must make sure that it is not an anti-reader search engine optimization, but must always make sure that it is, so to speak, also good and functional for the person who reads it, and that it is associated with a good experience. But here, too, I can imagine that there are certain automatisms that support the editor's work.

Benjamin Kolb [00:17:47]:

Now there is quite a lot in the area of audience development, which goes in the direction of personalization, automated A/B testing, optimization of some paywalls and so on. Do you see these as paths that lead in the right direction, where AI should be used to automate this, or do you also see a danger there?

Olaf Deininger [00:18:06]:

There is a large area that we see in various industries where AI is already able to implement relatively precise prediction systems. I can approach the issue by having an AI look, for example, at which user segments are most likely to sign up for a digital subscription. And then, in a second step, I can go here and say, "Okay, maybe I'll work on this potential in particular, or I'll pay special attention to this potential with my sales or marketing measures." That's where this can be helpful and, so to speak, also in the reverse view. I can use such prediction systems to show me which subscribers are particularly likely to cancel, and then I can also consider in a second step: "What can I do to ensure that they don't cancel after all?" Small, simple example: it is said that Netflix has determined that the probability of cancellation increases above average for Netflix users who use Netflix less than 15 hours per month. Now I can think, and Netflix did, "What do I do about those who are around the threshold and dropping? What could I possibly do to get their Netflix usage to go above that magic threshold?" That then led Netflix to optimize their recommendation system accordingly. And for things like that, I can already use AI today and maybe optimize my sales, marketing, my processes and so on.  

Benjamin Kolb [00:19:47]:

Exactly, you have now spoken of four AI methods that are in the area of statistics, i.e. probabilities, patterns, references, classifications. These are the typical mathematical basics in artificial intelligence. But now there are also deep-learning systems where, in the end, humans can no longer really understand the decision-making process, why the machine produces such results. There are also many who associate AI with the concern that machines make decisions that humans would not make and that are not at all comprehensible. You recently wrote about a book by Ferdinand von Schirach called "Jeder Mensch. How do you see it, since you just talked about recommender systems, i.e. suggestion systems? There's always the danger that you stay in your information bubble, that certain opinions that are out there or certain articles that work well are overemphasized, and as a result, even though they may not be true, alternative truths are created, because they are spread everywhere incredibly quickly. So how do you see that, the challenge of these artificial intelligence methods?

Olaf Deininger [00:20:55]:

This problem, that we may be influenced by algorithms that shape our media landscape and that increasingly - media also influence us more and more. The term "media" has also expanded quite a bit. Media" no longer just means the daily newspaper that I might still get, but in principle almost everything that we consume on screens, so to speak. And that's where the question of the bubble is really extremely relevant. And we haven't really got to grips with this socially, and we haven't really solved it yet, what happens to a society when people are only ever fed back a section of reality, so to speak. In order to be able to solve this problem - I know we're getting into a very, very abstract area here, but unfortunately, from my point of view, it's not easy with this topic - for this reason, we have to understand socially what AI is all about, without prejudice, and without all this weird hype about natural versus artificial intelligence. And once we understand what artificial intelligence really is, then we need to consider: How do the algorithms have to be so that we don't create a society where the individual finds himself more and more in a very individual media bubble and, so to speak, the common ground in a society automatically becomes smaller and smaller and, so to speak, the fault lines become larger and larger. And in order for me to be able to regulate something like that, in order for us to be able to deal with something like that, we need a realistic idea of artificial intelligence, and from my point of view, we are unfortunately still a bit far away from that in the discourse at the moment. And that's why I thought Ferdinand von Schirach was great for recognizing how extensively AI algorithms intervene in our society, not as classes of risk on a catastrophe level, but very subcutaneously in many small steps. And in his booklet, he calls for politics to deal with this, and in my view, this is really overdue at this point.

Christian Kallenberg [00:22:55]:

Olaf, now we've just discussed the challenges of AI a bit, but let's stay very briefly with a practical example. From your experience: Who among the large or small publishers in Germany or internationally is really using AI in an exemplary way?  

Olaf Deininger [00:23:16]:

I don't want to claim that I have an optimal overview of all AI projects that are taking place in German media or publishing houses. I really believe that there is hardly anyone who actually has this overview. But what I do know is that there are a whole series of content management systems that have quite interesting AI features. There are a number of daily newspaper publishers that are working on churn prevention, as we described, and there are a number of media companies that are working on identifying the potential of target groups that are particularly willing to take out a digital subscription. There are a lot of different projects, different statuses, but on the other hand, it has to be said, there are a lot of companies, especially in the trade publishing sector, that somehow think this is science fiction and that, I have to say, took a wrong turn a few years ago and today don't even have the opportunity to try out AI applications because they simply don't have the necessary, available digital data about their subscribers, customers or leads. And without digital data, of course, I can do relatively little with AI, at least actually. Unfortunately, I can only partially judge the situation in the area of content management systems, but it can be assumed that there are still many content management systems that are so outdated that they are not yet properly set up for this.  

Benjamin Kolb [00:24:49]:

If you take a look at the publishing landscape, you'll see that there are the big players, the corporations, which in principle are also very good at tackling this kind of thing with their own IT departments. What advice would you give, let's say, to a medium-sized publishing house or a regional newspaper publisher on how best to approach the issue?

Olaf Deininger [00:25:09]:

So the keyword, in my view, would practically be "the data-driven media company." As far as their subscribers are concerned, many publishers only have their name, address and bank details. And that would be the first step where I would start, that I would think about: How can I get reasonable customer data? How can I differentiate them? How can I develop it further? How can I define a path so that I can continue to qualify the customer data? Because that's exactly what we know from the advertising industry and from advertisers that in the future they will also increasingly want qualified, differentiated, digital data about target groups. And a second big area: When we talk about media companies, we always talk about content, of course. And that's the second point. I would look at - and this is also a very simple first step -: How is my content management, including digital asset management and so on, set up? Do I still have up-to-date systems? What could the requirements for the systems look like today and in a few years' time? And then I would look at whether what I have is still sufficient, whether I need to develop it further, what a path into the future might look like and how I can, so to speak, also keep my data digitally in a reasonable granularity and with a reasonable amount of metadata, so that I can possibly do more and more with it automatically in the future.

Benjamin Kolb [00:26:45]:

Let's assume that the company is now ready and has all the data. How many intermediate steps are there before it makes sense to attend your predictive analytics webinar?  

Olaf Deininger [00:26:55]:

So this webinar can be done without any prerequisites. I'm not giving any instructions on how to program or code something like this, but the goal of this webinar is to give the participants an impression of what it's all about, so that they get a differentiated idea of this topic, so to speak, rather on a conceptual level, so that they can then consider in their own company: "Where could this be relevant for us? And who could we talk to? Who would be an interesting contact? How could we adapt the topic, so to speak, in our house?" Or "adopt" might not be a bad term either. In other words, it's actually a bit of a beginner's webinar on this topic, which is intended to convey what you can do with it in the area of digital, but also print subscriptions and churn, and how it basically works.  

Benjamin Kolb [00:27:46]:

What do you expect a company like that, a medium-sized publishing house that now has its data, to do with it, so to speak, to use predictive analytics?  

Olaf Deininger [00:27:55]:

Marketing and sales are always associated with costs, and we know many industries - or that actually applies to every industry. If I practically bring my sales measures to the attention of people who won't buy a product or sign up for a subscription anyway, then that's actually burning money. In other words, my marketing or sales investments are most effective when they lead to a sale. And I don't have unlimited sales resources. And if a prediction system helps me to concentrate on the people where I have the highest conversion rate, then that is a cost saving that also has a very strong commercial effect. And in the other direction, of course. I also have the costs that arise in order to generate a subscription. Many publishers have already analyzed this quite well in the past and have the corresponding figures. That means I always have an investment that I have to put in to generate a subscription. If someone cancels, this investment is gone. If I can use AI to analyze in advance which factors lead to a cancellation, then I might have the option of turning off the factors. Let's take a very, very simple example: daily newspaper publishers have determined that poor delivery is a very, very strong cancellation driver. Now, you could say that you could come up with this without AI, but it's a wonderful, striking example that makes it clear that I can turn off a factor that basically causes problems by displaying the people who have complained perhaps three times in the last quarter about non-delivery of my newspaper, and then I can take a look at where the causes lie. Maybe it's the location, maybe it's the delivery person most of the time, maybe I can solve the problem if I change the delivery person. And as soon as I have this data available, so to speak, I can take action. But if I only learn about the problem in the abstract, that I only get a number from my call center perhaps once a month, that 18 people have complained again that the delivery has not been made, then I have no way of changing anything. From my point of view, this would also be a point where processes can be optimized, expenses can be reduced and investments can be protected.

Christian Kallenberg [00:30:20]:

Among other things, you advise publishers not only to use predictive analytics, but also to build up their own communities. As a publisher, I would naturally be interested to know why I should do this at all. What's the point? First and foremost, it costs money, doesn't it?

Olaf Deininger [00:30:37]:

Absolutely. A lot of what you do is associated with costs and investments. The approach is that we see in many places that readers and users are becoming more and more volatile. That is one development. At the same time, we see that the oversupply of media or the range of media on offer is constantly increasing. That means, so to speak, that the user is becoming more volatile, the supply is increasing, and thus competition is also becoming tougher. At the same time, the fourth factor is that distribution pressure is becoming less and less effective. In the past, in the 1950s or 1960s or 1970s, marketing budgets were practically ramped up, market pressure was practically built up, and this compensated for it or enabled it to be controlled quite well. And that is also something that is working less and less. Today, I can no longer successfully market a product by simply developing a great deal of advertising pressure. It's still a factor, but it's no longer as effective as we saw it 20 years ago. Media companies are faced with the challenge of having to think: How do we engage our users and our readers perhaps beyond the media we've been producing for some time? And here, of course, a community could be a great solution, because many people are looking for a digital home, even in the specialist field. Many readers, many users are interested in exchanging ideas with other readers, users, experts and so on, also with the editors. And now we are in the situation that we have the technical possibilities for this. Either I can solve such an issue via platforms. A very simple solution would be a Facebook group, for example, but I can also go one step further and say: "I'm not building this house on someone else's property, where Facebook might then at some point come up with the idea of building a four-lane road through there, but I'm building my house on my own property and perhaps also establish my own community in my own branding via a white-label platform solution and then try to use this to balance out these four tendencies that I described at the beginning accordingly, create yet another additional platform for how I can bind users or readers to my medium with further added value." And if you look at what these platform solutions, these white label solutions cost in the meantime, then we are so like WordPress also at a very low level. In principle, that's not rocket science anymore, it's commodity. And so the investment required to build an online community for my online medium is no longer so insanely high, which in my view makes it an effective means of absorbing this volatility.

Benjamin Kolb [00:33:20]:

Yes, this is actually a very exciting area, i.e. own platform versus third-party platform. One basis is the data that you have when you have your own platform. That's a big advantage, in that you have the user data directly and know what the usage behavior is like on the platform. I also believe that when you're out and about on other people's private platforms, you naturally have the advantage that they already have mechanisms for attracting customers to my platform. They're all already networked with each other on Facebook, which means it's easier or the hurdle is lower for me to get new members if I have an exciting topic. Of course, that's something you have to weigh up. A few days ago, we read that - speaking of external platforms - Axel Springer has also decided to rely completely on Facebook News, surprisingly enough. How do you see the use of foreign platforms for media companies? Is that something they shouldn't do, because they're giving away their entire data landscape, or what's your opinion on that?

Olaf Deininger [00:34:17]:

So here again, unfortunately, a very differentiated answer is possible. There are a number of platforms that have a very high proportion of social media reach in their traffic. In some cases, up to 30, 40, 50 percent of their reach, which they then practically monetize in the advertising market, is generated by social media platforms. For them, of course, cooperation with the relevant platforms is extremely relevant. And we've also seen in the past that when Facebook suddenly changes its algorithm without notice, 10 percent of the reach is quickly gone, which in turn can be a problem in the advertising market, inventory and so on, because customers call and say, "Tell me, there's no more reach coming from you. Or the reach that's coming from you guys, that we have firmly budgeted for in terms of sales, it's dropped 10 percent. What's the reason for that?" In principle, that means: on the one hand, these platforms are part of the solution for media companies, but on the other hand, of course, they are also part of the problem because, as described, they can change the rules overnight and that has negative effects for the media companies' media business. That's one way of looking at it. There's another consideration. If I'm perhaps a small trade magazine and I say, "I'd like to get some initial experience of what it's like to build an online community," then of course a Facebook group that I configure within three quarters of an hour can be a very nice solution. Little investment, low time-to-market, quick to implement, quick to administer. So I'm sort of, if you want to put it that way, quickly in the game actually and so that would be part of the solution again. Unfortunately, there is also a counterexample. The New York Times has built up a cooking community on its own website as a sub-domain, and has had a Facebook cooking community upstream, which, I believe, with over 70,000 members, took on the form a few months ago with many discussions, arguments, and negative communication that the New York Times has announced that it will withdraw from this Facebook group, and that it will, so to speak, release it, and that it can no longer look after it with its own moderators, because it would need more moderators to look after it properly. They don't want to make this investment, so they are withdrawing from the platform. So in principle, the platform has failed a bit because of its size. That's the reverse side of the coin, so to speak. Bottom line - what can you say? First of all, I think every media company has to make its own experiences with the issue of community. There are tools that allow you to do this without much risk. I would recommend this to everyone, so that they can then decide later, so to speak, on the basis of this accumulated knowledge: To what extent do we get involved, do we invest? At the moment, I know of many media companies that want to charge for their existing online communities. Here, too, the possibility of monetization arises, and there are already platform solutions where I can not only build my own community with my own functionalities, which I can design with my own branding, but where the monetization tools are also attached.

Christian Kallenberg [00:37:39]:

But from your expertise: Who takes care of the communities? As a medium-sized publisher, do I need my own community manager for that, or is that something that the editorial team helps with?

Olaf Deininger [00:37:51]:

I know editorial offices and editors and colleagues who have a lot of fun exchanging ideas with their readers, discussing with them, asking them questions, putting forward theses, letting them discuss. I also know of platforms that run out of the editorial office as - I don't want to say "fun project," but almost as a fun project. Of course, it's work that works without a separate moderator and brings a lot of joy to everyone involved. On the other hand, I also know editorial teams with editors who have no fun at all with something like this, and if you then force them to do community service for three hours on Friday afternoons - I don't want to paint too negative a picture here - then you might find that their commitment to the community is perhaps average at best, which in turn suggests that if I have a constellation like that, you should install a moderator who will ultimately take care of this community. Of course, it also depends on the size of the community. What I find quite charming is when the editorial team itself enjoys and develops a sense of fun in exchanging ideas with its readers. I think that's part of editorial work. It's quite normal. It's not really worth talking about. It's a matter of course. But I also know that there are colleagues and editorial teams who see things quite differently. That may be legitimate. For the future, I think that we as journalists and editors cannot engage intensively enough with our readers, users and target group, because this is where we are most likely to create the opportunity to learn something new, to learn where their pain points are, what they are interested in, where they want to go, how we can further optimize our information offering so that it becomes even more relevant for our target groups.  

Benjamin Kolb [00:39:42]:

What are your experiences with artificial intelligence methods in the field of communities? For example, recognition of trolls and negative posts, sentiment analysis? Do you have any experience that you can pass on to publishers who are starting a community in order to perhaps simplify the life of the moderators?  

Olaf Deininger [00:40:00]:

So what I'm noticing is - has been for 10 years, actually: The fear of trolls is regularly overestimated. As far as technology is concerned, I have to say that my experience is very limited. Quite honestly, I can't think of any real application that could be recommended at this point. You can certainly transfer all the techniques that we've had here from the area of prediction systems, pattern recognition and so on, but I think that's an area where things are really just getting started. And there are many things that we don't know, or don't know yet, or only know in part, or are such that they are perhaps not even recommendable.  

Christian Kallenberg [00:40:42]:

Olaf, let's get back to the fun part of the job. You started your career in journalism by getting into debt. Is that right?

Olaf Deininger [00:40:52]:

Yes, you can say that. It was a long, long time ago. From today's perspective, it was an extremely manageable amount, but I paid the first printing bill for my first school newspaper myself. And because as a young person, as a teenager, you didn't necessarily automatically go to the cheapest print shop, and perhaps you weren't able to estimate prices at the time, you ended up with a bill that exceeded my pocket money budget by quite a bit, which I then somehow paid with family help anyway.  

Christian Kallenberg [00:41:29]:

By now you know your way around journalism and you have a second side project, the topic of food. I've already noticed, Benni, that somehow the topic of food is very preoccupied here in this podcast with all the guests, except me - I just like to eat. Benni, you are a hobby cook, Olaf is also a hobby cook, or even the main person responsible for the topic of food at home. Olaf, what is your favorite recipe at home now?  

Olaf Deininger [00:41:57]:

So it's right. In the household with my wife, I am basically responsible for the hot meals. And if I don't take care of that, then there are no hot meals. So every day, like every househusband, I am faced with the question: What am I going to cook today? And that's why I love recipes that produce the maximum effect with the minimum amount of time and resources. And of course, like every househusband, I also make sure that the range of products is as wide as possible, so as not to promote a one-sided diet. So what do I do relatively often? Salade niçoise is something I'm making relatively often at the moment, especially when it's not raining all day, but when it's warm and spring-like outside, but I've also been making shakshuka relatively often lately.  

Christian Kallenberg [00:42:50]:

What is it?

Olaf Deininger [00:42:51]:

Shakshuka is such an Israeli dish or a dish from the Levant. It's basically peppers stewed well with tomatoes in such a good sauce, enriched with a relatively large amount of cumin, and basically an egg beaten into it, but in such a way that only the egg white sets, the yolk is still liquid. And then serve that in the consistency, so to speak. It's a really great dish. It's great with white bread or with pita bread.  

Benjamin Kolb [00:43:21]:

Sounds excellent.  

Christian Kallenberg [00:43:22]:

I know we could go on talking for a really long time, but I'm hungry. That means you've made my mouth water now. Olaf, thank you very, very much for your time. Maybe we can post your recipe in the show notes so people can take something practical away with them. No, fun.

Olaf Deininger [00:43:36]:

No, we're happy to do that.  

Christian Kallenberg [00:43:37]:

Very good. And of course, maybe one or two other links of good examples that you mentioned. And otherwise I would say Benni, good episode so far?  

Benjamin Kolb [00:43:46]:

Yes. Olaf, thank you very much.

Christian Kallenberg [00:43:48]:

And if you, dear listeners, have any feedback on this episode or any requests we should discuss in one of the next episodes, please feel free to send us an email at We'll be in touch. Thanks for listening and see you next time.