How to successfully build a premium offer
4th Episode Audience Development Deep Dive
We Like Mags and SPRYLAB in conversation with André Lux, Head of Sales & Marketing at Heise Medien
Heise Medien publishes the industry magazines c't and iX, Technology Review and Telepolis, among others. Since 2019, Heise Medien has been offering the subscription service Heise +. For an additional fee, interested readers can delve deeper into industry topics than was previously the case via the free portal Heise Online. The maxim here is: "We are deep divers and get to the bottom of things. To this end, a dedicated online editorial team was created to deliver just that to the target group: Articles and background reports with even more depth.
Price juggling not without consequences
The pricing was still bumpy at the beginning: Interested subscribers had the option of paying a temporary price of 9.95 euros, which was to increase to 14.95 euros at a later date. However, this tended to cause confusion among readers, so Heise Medien quickly switched to a fixed price of 9.95 euros. But that was not to be the end of the story: When the prices were increased by three euros in March 2021, the company quickly felt the reactions. Digital subscriptions plummeted and Heise Medien realized that online offerings have a different sensibility than the print market.
André Lux also reports on the intersection of print and digital subscribers and the migration from print to digital and vice versa. This is a challenge that the company must face if it wants to continue to consistently retain and build up readers. In addition, the online editorial team of heise+ works with a topic radar that provides the editorial team with important information about relevant and actually interesting core topics in order to reach users.
But that's not all: in order to retain new customers, their on-site behavior is measured, albeit not on an individualized basis, but in "mini-cohorts. According to André Lux, the aim is to offer Next Best Content (NBC). This is content that is so exciting that interested parties accept the Next Best Offer (NBO) in the next step in order to secure access to the exclusive content. Currently, heise+ has 46,500 active subscribers.
What awaits you in this episode
00:45 - The big difference: heise and heise+
04:06 - Subscription and pricing: why the online market is more sensitive than the print segment
09:11 - The difference between One- and Multibrander
15:25 - Collaboration between marketing and tech team
17:11 - Print versus digital subscribers
19:48 - Why a topic radar is useful
26:21 - Customer Audience Management: How to nurture and woo customers
29:41 - Open sesame not or how Heise + works with paid and gated content
33:49 - Newsletter Marketing
36:05 - A/B Testing: What Works and What Doesn't
38:37 - Traffic - Mobile or Desktop
40:21 - Customer Journey & Warning Signals
44:47 - Acquisition t3n: What follows the takeover of the successful industry magazine.
46:39 - André Lux private
Christian Kallenberg [00:00:07]:
Welcome to the new episode of the Audience Development Deep Dive from SPRYLAB Technologies and We Like Mags. Also sitting across from me today as co-host is Benjamin Kolb, CEO of SPRYLAB Technologies. Hello Benni.
Benjamin Kolb [00:00:20]:
Hello Christian. I am very happy.
Christian Kallenberg [00:00:22]:
I'm also looking forward, above all, to the following conversation with André Lux, Head of Sales and Marketing at Heise Medien. Hello, André.
André Lux [00:00:30]:
Hello Christian. I am glad that I can participate here today.
Christian Kallenberg [00:00:34]:
Today we want to know a little bit about what's going on at heise+. André, would you like to tell us something? How long has heise+ actually been around? What is it actually? What can it do? How many people want to know about it?
André Lux [00:00:45]:
Yes, so heise+ is our flat offer that we have now launched since the end of January 2019. The offer mainly includes in-depth articles that are explicitly written for heise+. So this doesn't affect the news that has always been free, i.e. via heise online. Those are still available just as freely as they were offered back then before the end of January 2019 with the product launch. What is new is that an extra online editorial team has been created that is precisely for our target group - for what c't was and is already known for and iX, that we simply look even deeper. We've actually always had the phrase: We're not the snorkelers, we're the divers, the deep divers, and we get to the bottom of things. And that is ultimately what the heise+ editorial team does. In addition, and this was of course totally helpful and useful at the beginning, it is also a flat offer to all our magazines that are published here under Heise Medien. And we had hoped that this would enable us to reach a very broad target group, from a beginner or a hobbyist, from Make: all the way to iX, where the professionals are on the move, the admins, with content that I no longer understand.
Christian Kallenberg [00:02:12]:
Okay, that's a relatively heterogeneous target group. What does that look like? How many subscribers are there at the moment?
André Lux [00:02:19]:
We currently have around 46,500 active subscribers who use the product, and we still have a breakdown. I can explain that right now - in case it's not a question. On the one hand, of course, we have new customers who are completely new to this digital product, and about half of these 46,500 are represented by them. And the other half are actually existing customers whom we have made an upgrade offer to. We have also differentiated between the two. Let's say a "multiple subscriber", i.e. someone who has two or more subscriptions with us, and there are a few of them, they get heise+ for free. So they have free access to heise+.
Christian Kallenberg [00:03:05]:
André, very briefly.
André Lux [00:03:06]:
Christian Kallenberg [00:03:06]:
When you say, "Two subscriptions with you," that's two print subscriptions in that case, right?
André Lux [00:03:10]:
Yes, or it can also be a purely digital subscription, but from the print brand.
Christian Kallenberg [00:03:13]:
André Lux [00:03:14]:
Exactly, and then there was another option, if someone only has a subscription with us, i.e. standard subscription, then they got an offer of €5 extra payment per month. And the other option was that he had to pay 3€ extra if he had a Plus subscription with us. It sounds silly now, heise+ and Plus subscription, but we have always had - this has been the case for decades - with c't, for example, and with the other objects as well, a Plus variant where you could and can access the archive and where you can also access the digital content, i.e. read browser-based or read as PDF.
Benjamin Kolb [00:03:51]:
On the subject of pricing, perhaps one thing: The prices were probably set in advance. And did you do something like market research, or how did you come up with the prices? And are they set in stone for you now, or would you like to deal with these prices more dynamically in the future?
André Lux [00:04:06]:
So the start was relatively easy. When we started out, it was customary in the market to pay €9.95 or €9.99 per month. And we adopted that and said that was actually the right price. We actually had another Berlin agency on board that advised us, and with them we finally came to the conclusion: "No, we should set an anchor price." The anchor price was 14.95€. We wanted to communicate this price throughout. That's exactly what we did at the beginning, saying: "Watch out, now first the special price is 9.95€. It will then cost €14.95 at some point." That wasn't quite so clever, because it naturally prompted a lot of questions from people who wanted to know: "What does that mean exactly? Is it next year, the year after next? When can I pay €14.95?" That doesn't help people sign up quickly. And then we said, "Okay, we'll say goodbye to the €14.95 for now, our actual anchor price, and then actually just go to the €9.95." And that's what we decided after two weeks, that we would set our new price at €9.95 and then leave this anchor price and cancel it. In the course of that, however, we then also said, "Okay, of course we also want to implement price increases," and that's exactly what we did at the end of March 2021 this year. We raised the price to €12.95 and then we also noticed that clearly. So we didn't do a test before that either. We then said, "Okay, we can do it like that. And the price elasticity will probably be similar to that of our classic products, i.e. print products," but we were actually very wrong about that. So there is a completely different sensitivity in the market. We also felt that. We then noticed a significant drop in orders. That's when we said, "Okay, we have to act, we have to react." We didn't actually do a lot of price tests in the past either, and that suddenly became an issue that we said, "We have to introduce some kind of mini-subscription or event subscription or seasonal subscription, whatever there is." And the variant that we then chose was a summer subscription. That was then again at half. So 12.95€, half was then 6.45€ for us. And then we followed through with this offer to get people or potential customers, the users, to subscribe to us again. And that actually worked well. We're now looking at the cohorts to see what the conversion is like afterwards. In the course of this - and we offered this to our existing customers - we said: "Watch this. You as an existing customer, you don't have to pay the €12.95 now, but you'll continue to stay at your old price of €9.95, but that's only if you decide to take out the annual subscription." So, we actually sent out test mailings to 1,000 customers and made various offers and were actually able to get a total of 14,000 existing customers to sign up for the €9.95 annual subscription. However, we also had a cancellation rate of 8%. That also hurt us. But we realized that an annual subscription works. And in the meantime, we have actually implemented this as a normal form of offer in the checkout, i.e. a curtain.
Benjamin Kolb [00:07:28]:
So actually an A/B test by mailing.
André Lux [00:07:30]:
Benjamin Kolb [00:07:30]:
So to speak. Are there any plans to do something like this directly online, i.e. to set different prices for certain cohorts, which various publishers are already experimenting with a bit, but which is of course always a bit strange for the user when he knows that others might get different price points?
André Lux [00:07:50]:
Exactly, so we've had the idea for a while. I presented it sometime in the summer, that we want to operate via a smart paywall. In other words, the idea behind it was that we simply look at segments: How does someone behave on the page? Is this a fan, a hardcore user who comes every day and stays with us for an hour every day? That's just an example. There's probably no such thing. Maybe there are, but I don't know how many. Or is it someone who comes three times a week and reads for ten minutes? Or is it someone who comes once a month and reads for fifteen minutes? So, and of course you can treat them differently, because the hardcore user, he's probably more willing to pay a higher price. And those who surf by once in a while, but find us quite good, may be willing to pay anyway, but then at a lower price. And we've actually talked about this, but we haven't implemented it yet because we didn't have the time and resources to develop this further, to build this smart paywall. We still have that in the back of our minds, we want to do that, but it's not current at the moment.
Christian Kallenberg [00:08:57]:
Then let's talk again about another number that's already in the room, which is the 100,000 subscribers.
André Lux [00:09:02]:
Christian Kallenberg [00:09:02]:
Which you are aiming for with heise+. You presented this some time ago at the VDZ Publishers' Summit. How did you get there? And above all: When?
André Lux [00:09:11]:
We will manage that, for sure. We will also achieve this in 2025. We have launched various initiatives in our company, and one initiative is paying a great deal of attention to editorial work, to ensuring that we write more articles of the quality that customers are used to. That is a very important initiative that we have launched. Another, second important initiative, which will then come - it will now come successively - so in January we want to start it properly and have it accompanied here in house by external partners who are experienced in implementing such things. It's really a matter of trying to set up teams along the customer journeys and saying, "Okay, it's relevant that we focus on the acquisition environment, the subscription environment, and the customer loyalty environment. We often lack focus here. On the one hand, this has to do with the fact that we are not a one-brander. We're not Spiegel or FAZ, where everything is focused on one brand in order to generate the greatest possible success. Sometimes that makes it easier, because that's the product and you simply have to be effective. We have a lot of products. We have also just launched - "Silent Launch" is the name of it with the Academy - heise Academy. This is a lifelong learning program. You can also take out a subscription. Businesses can do that, and so can private individuals, i.e. B2C. But what's exciting for us here now is that we're saying that we're setting up teams along the customer journey that are interdisciplinary, that ensure that, for example, in audience development - that's heise online, so to speak, not the whole route, but heise Online - that we say there: "We want to optimize the reach, we want to increase the visitors per month, we want to increase the social media reach, we want to increase the social media entries, we want to increase the number of newsletter recipients." So, there has to be a focus on that and there are teams that are put together from editorial, from marketing, from the Data Competence Center, from technology, product management, in order to simply get this view of this section within the customer journey. And it's the same with subscriptions. There are simple questions in onboarding: How do we have to do sensible onboarding? How individual do we have to make it? And when does a subscription that is purchased in this context actually become a subscription? That's where you can now set time windows that the onboarding has to be so good that it stays on board for at least 60 days, 30 days, 90 days, and only then does it go into the inventory and then it's in customer retention or engagement, where we then look to see that the heise+ users actually use their subscription - but we'll probably get deeper into that later - to see that we reduce the churn rate, increase the win-back rate, and simply improve the reading behavior, because the new thing and the thing that we're also learning right now is: The digital customers we have, compared to our print customers - and even there the print customers who have digital - they behave completely differently and they no longer get a magazine somehow in the mailbox and have an event about which they are happy, but they simply have to be triggered by us at the right individual times.
Benjamin Kolb [00:12:33]:
That's super interesting with the teams along the value chain. My question would be, because we are also moving in the technical area with our own products: You said that the teams also include technicians, for example, and probably also developers. How deeply do they intervene in your system landscape, so to speak, with tools that provide assistance or even go in the direction of fully automating some processes? Or are they still in the status at the moment that they are more active in an advisory capacity, because in principle they gain knowledge and provide assistance?
André Lux [00:13:05]:
I'll probably be able to give you a final answer at the end of next year, because we're currently on the way and want to start the journey, but what we want to achieve is that within each block in which you're on the road on the customer journey or responsible for the team, the colleagues work out the measures together: What is relevant now in order to achieve the KPIs that I have more or less just named - you can also put numbers behind them - in order to get there? And then they will say: "Okay, there are technical approaches to solutions. There are tools on the market that can be used. We have to act accordingly in marketing. The editorial team has to be involved." And for me, it's really a matter of thinking together within the section in which we move along the customer journey, and then finding a commitment: How do we want to achieve this together? And the colleagues who are in the teams, so to speak, remain in line. They remain attached to the web devs or remain with product management. And it's important that there is a commitment along the line to say: "Okay, we're working positively to ensure that what was developed there is also implemented accordingly," but that doesn't mean that this is now a fast track, so to speak. And the topics that we continue to have here - Academy, classic products, verticals that we have - these must continue to take place and they need just as much attention and the resource that is used there. But then you have to decide on the measures that are developed: What has the highest priority for the company? What works best for heise+, too, if the 100,000 are to be reached? And that is, so to speak, where we want to move. Once to create this spirit, this "pulling together", and to have the permeability of knowledge in every area and that a technician also knows: "Why does marketing think the way it does?", or we also better understand: "Why is it sometimes not fast enough with technology?" It's never fast enough, but they have their reasons why what works the way it does. And that's what we're trying to achieve, so that there's a better understanding and also a better implementation of what you want to implement within this block, so to speak, or of what the team wants to implement or has to implement.
Christian Kallenberg [00:15:25]:
I could also imagine, André, that it might be a bit easier in a publishing house like yours, which also deals with a lot of technical topics in terms of content, that the cooperation between the individual teams works better than perhaps in a publishing house where it's primarily about handicrafts.
André Lux [00:15:41]:
I don't know what it's like in a publishing house that deals with handicrafts, so I can only talk about what it's like here. But, yes, of course they know what's available, and they already use a lot. They use our map tools, the SISTRIX. They use everything that we generate in terms of values via our Customer Audience Management - we call it "CAM" here. They know which A/B tests we do, they are also involved in this project. This means that we don't actually do anything without editorial staff, but that's also absolutely necessary, because ultimately transparency should be there: What can technology do? Where does technology help me? Where does it support me? And where does it make me better? And that's what it's all about. Better, more efficient, more effective. Depending on what it is. And I believe that this, of course, must also continue to arrive here. That's always a bit of a challenge - to trust it in the first place. If they think: "But I would have produced more subscriptions if I had put my articles there, and now CAM has put someone else there. He's producing less somehow." You don't know, because the other one wasn't there, but that's also A/B testing. We have already done that. Sometimes it swung left, sometimes right. There was not the right decision. In this respect: We are actually still here, as far as this is concerned, this measuring, testing the algorithms - it is actually the case that we are still on the journey. The system is learning every day, and you have to let it learn. If this learning of the algorithm is not made possible, then it will be difficult to achieve the results that we all want.
Christian Kallenberg [00:17:11]:
Before we talk a bit more about CAM, let me take a step back. I'd like to know once again what the intersections are between your print and digital subscribers. You briefly mentioned earlier that they are completely different pairs of shoes. Are there still people who switch from one offer to the other? Or, if they do, is it just a supplement to the offer they originally came in with?
André Lux [00:17:36]:
No, so that's of course what we tried to do with the pricing that I mentioned at the beginning - that we try to design it in such a way or have designed it in such a way that we don't want to offer any loopholes, so that someone is a Mac & i customer - we love our Mac & i subscriptions, but they pay less in an annual subscription than a heise+ subscription costs, so to speak. And with that in mind, we didn't want anyone to be able to slip through the back door and read the heise+ content. But what we do see, and this is completely normal, and this is also what the classic editorial team or editors have always said: There is a latent danger that they will migrate away, because if they want to read digitally, they will order heise+ and then they won't order our individual brand if they can read all brands. That can actually be emphasized to a certain extent. We are already noticing with our traditional brands that the digital share of subscribers, which has always risen over the years, is currently moving sideways. That's the case. We are actually also seeing migration away from the classic brands to heise+. So we have a total of about 6,000 subscribers to c't, for example, over the last one and a half or two years, who have actually canceled c't and said, "I'm only going to use heise+ now. We don't know exactly, because we haven't measured them yet either: Is this a departure in installments, that they first say, "Okay, c't, I don't want print anymore and now I'll take heise+ again and then I'll get out completely after three months," or have they found themselves in our cosmos, so to speak, and is that the product with the variety, with the somewhat different content orientation, that then appeals to them, so to speak, and that they also keep in the longer term and are willing to pay for it? I can't answer that here. We are still in the process of collecting these figures, but for every object that is also represented in the flat, so to speak, at heise+, there are departures that have to be recorded. That is definitely the case.
Benjamin Kolb [00:19:48]:
One point that I still remember now, also from your presentation, is the topic of topic radar. Maybe we'll start there with the whole topic of how you want to position yourselves. The topic radar should actually determine the topics on which content is ultimately produced and how the audience is addressed, so to speak. What are your ideas and where do you currently stand?
André Lux [00:20:11]:
The topic radar goes in different directions. On the one hand, the topic radar is also intended to help the editorial team to see what our top topics are. So what are our users most interested in? And there we have, I'll give you a big thumbs up, 30 top topics that have been identified, that we know about: These are core topics that we serve and that are interesting for our users. What we also see about that is, are we over- or under-performing in that topic area or about that topic? That is, do we actually have many more PIs and can't provide enough content, or do we actually have far too much content and there are hardly any PIs that run on it? And an editorial team can then decide: "Okay, I can shift the focus. I can see where there is a need in topics, and I can write more about that. In this way, I can help ensure that we have articles in topics where we then also have a chance of converting to heise+." In other words, that's one of the topics we have to identify. The other topic area that we have: We also try to crawl a little bit. That means we simply look nationally, internationally: Which IT side is actually talking about what? And we make this available to the editorial team so that they can decide: Is there anything in there that is somehow flippantly the hot shit that should definitely be brought up? Or do we have it all on the screen anyway and there's nothing new? So just to have an overview of what's being published elsewhere, so to speak, and whether that's of any use to us if we keep an eye on it. Furthermore, we also have the opportunity in this context for our sales colleagues, I'll say in the advertising area - in our case it's Sales and Solutions - that they can of course also sell these topics to industry customers via the topics. So if someone is only interested in security or only in DevOps or whatever, then they can simply say, "Here, you can only book security," and all those who are relevant to this topic area, so to speak, are shown that, and you can then charge a higher CPM for that. So that's the idea behind it. That's what we do via the topics.
Benjamin Kolb [00:22:32]:
When it comes to crawling, I'd be very interested to know what exactly you do. For example, we use publicly available news feeds, social engagement data, and trend data that can be accessed in order to learn what is exciting for the editorial team in as automated a way as possible, to cluster topics based on the content we produce ourselves, and to assign them automatically, so that the editorial team receives suggestions for specific departments as to which topics are currently on the rise or which should be on the seasonal agenda again, in order to relieve the editorial team of as much of the manual work as possible. What do you do there?
André Lux [00:23:15]:
In fact, technically under the hood, I'm a bit weak in the chest. I admit that in all honesty. We have colleagues here who can do it much, much better, who could explain it much better than I can. I only know that we have two colleagues here, one colleague and one colleague - the colleague who created this topic model and also created and presented this radar for the first time. The crawling is now done by a new colleague that we have acquired, who is pursuing the topic. How we actually cluster this and what we can and already do or want to make available to the editorial offices, I can't really say at the moment and that would be the crystal ball that I'm taking out, so to speak. I actually don't know. So I know we can, we do it for us already. What the editors already use of it, I think, is very little. They probably know that we're doing it, but I don't think it's in the processing quality yet that the editors can work with it yet.
Christian Kallenberg [00:24:13]:
André, I think it's very, very remarkable how deeply you know your stuff anyway, because if I've read your CV correctly, you actually come from the subscription sector, and I assume that in the past it was mainly print subscriptions that you dealt with at the beginning of your career. How did you get so deeply involved in that now? Do you have an affinity for technology in your private life, or have you read up on everything at heise in the years since you've been there? Well, I think that's pretty impressive.
André Lux [00:24:41]:
Well, I only read c't and iX. No, I don't. I don't. No, the fact is that for a while I was still at Bauer Verlag in Hamburg. That's now Bauer Media. And at some point I was in charge of direct marketing, I think it was called, but before that I had already set up the online store, and there was no Shopware 6 or 5 or whatever there was - no Magento and so on and so forth. They actually did it with someone else and customized the store for us. In the course of that, I also did a lot of e-mail marketing, which meant that we had response rates. There is no such thing anymore. And the word "spam" didn't exist back then either. I think I've always been deeply involved in technology, because I'm interested in it and because I try to at least sort out, evaluate and see everything that helps of a technical nature in marketing or in the overall constellation along the customer journey, everything that exists and everything that's new: What could be useful for us? Since we here at the company - and that's just the way it is at an IT publishing house - are very heavily involved with these topics and the Data Competence Center is also located in the department I manage, it's simply the case that we have to deal with these topics on a daily basis. All colleagues who work in this context know much, much more than I do. And that's good and right. It is inevitable that you simply want to have your say and also have to form an opinion on the matter in order to make a final decision or to pass it on to GL. That simply helps us to be more or less up to date.
Benjamin Kolb [00:26:21]:
One question that I'm very interested in is: You talked about the CAM, the Customer Acquisition Management module at your company, or maybe that's also a tool. Maybe you could tell us a little bit more about how that works and, above all, maybe also where it attacks and up to where the CAM area goes, so to speak.
André Lux [00:26:37]:
Yes, with pleasure. For us, it's always called "Customer Audience Management. For us, this is very much based on the topics of NBO, or Next Best Offer, and NBC. The whole thing is based on customers' on-site behavior. In other words, we measure customers as far as possible. We don't do it on an individualized basis, but we then make it anonymous, so to speak, and form mini-cohorts in order to have that reasonably clean. And the goal we want to achieve with this has always been to increase stickiness, so that when someone comes to heise online, they are happy to come back - to increase the shelf life within the subscription, i.e. with heise+. If we have recruited someone, then we want to offer him this - NBC it is, Next Best Content - so that he can read the right content, the content that is exciting for him, and of course we want to use this to generate sales, that is, the NBO, Next Best Offer. We try to do this in the best possible way using the data that we have about the users on site, so that we can form clusters. And we have formed the clusters. We then added purchasing behavior data, and then we have our DMP, i.e., we have on-site data, and if we identify the user when he logs in, we also have his purchase history, so to speak. We bring this together and then use it to calculate on the fly whether the person has a high affinity, a medium affinity or a low affinity, now especially with heise+. And if he has a low affinity, then we don't approach him, we would keep this potential free for other products and play it out there, so to speak. If it has a medium or high affinity, then it will be played. But we also found that the medium affinity - where we can probably offer other products even better - and the high affinity is actually where we then play it off, so to speak. This model that we have carried out here, we have tested it, so to speak, and it is correct. So that means we tested it through A/B tests and saw: Okay, when we do this, we are definitely better, we have better results. And we can then use this to play out the right advertising media, so that the low hanging fruit, so to speak - it sounds silly, but it goes a bit in that direction. And it's the same with content. If we notice that people have an affinity for topics, then we can play content to them. We have kept our clusters or segments, depending on what you want to call them, relatively rough. I think we have five or six segments that we sort into, so to speak. So after three clicks, we know which cluster we're going to sort it into, and then it's recorded accordingly. But we definitely want to refine that. We want to become more granular in order to get even better results. And our colleagues in the Data Competence environment are working on precisely these topics to make the whole thing even more effective for us.
Benjamin Kolb [00:29:41]:
Now, if I understand correctly, you have gated content, which requires registration, so to speak, and paid content, which then requires the subscription model in the end, right?
André Lux [00:29:52]:
Benjamin Kolb [00:29:53]:
Exactly, and that works very successfully for most of our customers. And this "guiding the customer, so to speak, into this gated content at the various levels", you also do this via recommendations, which means that for an article that I have read, further articles are then selected, so to speak, depending on the segment, depending on the cluster, whether it is opportune to now offer him, for example, an article that requires registration, because you expect more conversion from this.
André Lux [00:30:22]:
Exactly, so we are actually not yet doing that to the extent that we would like. We are doing it in some areas of the homepage. This has to do with the fact that we are in the process of converting our homepage to a - we always call it a "module system". Up to now we have had a relatively inflexible system and as soon as we have this module system, we will be in a position to use every template dynamically. Of course, we will make even greater use of this - of course, everything in consultation with the editorial team, that this makes sense and that we achieve the best possible effect through this, but that is the point of doing exactly that, and that is also the point of what I called "stickiness", that we, if someone has come in via an article - they are supposed to return to heise online - that we then offer them the articles, so to speak, which is what an Outbrain or something does via Recos, that we offer them the best possible thing so that they feel comfortable and then go through the paywall as smoothly as possible. And as I said, when he has gone through it, it should of course be exactly the same afterwards in terms of customer loyalty or engagement, so that he finds the right articles for himself, without wanting to form a bubble.
Benjamin Kolb [00:31:33]:
How deep have you already gone into this topic of the conception of such a recommendation? I remember: From our own experience, you can use as many parameters as you like to play out the best possible content - be it interests, or of course the content itself. They have their own conversion rates, the articles that you play out there. So if you have someone who, for example, already has a great affinity for the topic, and you think, based on the cohort assignment, that they might be about to register - do you then take all these factors into account in order to play out the piece of content that would convert them next or -
André Lux [00:32:09]:
That is the idea. That's not where we are at all yet. Most of the content, most of the articles, they just come through the editorial department and are played on. We have areas where we are doing exactly that now to achieve precisely this effect. Of course, this is always a bit of sporting ambition on the part of the editorial team. So on our part - we always think together, but there is such a sporting ambition: What actually works better there and how? But it is actually the case that we are also still in the process of determining: What actually leads to a high probability of purchase or reading? What are the attributes that actually contribute to this? And what are we missing? Because we see: Okay, we have clusters and we see that they react better and we have calculated affinities high, medium and low, but what exactly is it that leads to a higher purchase or reading behavior or probability? What attributes? And that's actually where we're still working on determining that as well, just to get even better. But as I said, currently, if you look at the homepage, you'll also see that a lot of what we're doing about dynamic playout is not happening yet. We have designated areas for that that we've agreed to just learn. And that's where we are right now in that status.
Benjamin Kolb [00:33:20]:
And if you now imagine the final expansion stage of this dynamic page: How flexible could it be in the end? In our case, it's basically done via cohorts, where you can then design the entire page dynamically with playouts, including whether the subscriber sees a subscription at all in this case, so in principle everything can be designed dynamically based on the cohort. Is that your plan as well, or will that essentially be limited to content, the dynamics?
André Lux [00:33:49]:
So in the main it will be limited to content, but nevertheless, when we talk about NBO, of course we always try to offer the best possible deal, because as I said, increasing customer value is also very, very high on our agenda. We simply look at: How can we get there so that we can design exactly this calculation to such an extent? And how can we work together with the editorial team to ensure that both are happy to play this? As a salesperson and technology enthusiast - I'm exaggerating here - I would say: Yes, we once had an event at our technology partner's, I don't know, five years ago. They told us about a customer they had. BABY! That's Vice. They reported, "Yeah, they have 225,000 newsletter recipients and they sent out 225,000 individual newsletters." That means each newsletter recipient actually got an individual newsletter based on their needs. I always have that picture in my mind a little bit. We have after, I don't know, our 100,000 heise+ subscribers and we then have 100,000 different heise+ products. It's all the same, because the content is there, but everyone gets it played up differently for them. And that's an idea I have in my head, but it doesn't have to match 100% with what the editorial team thinks is good. But the fact that we have the opportunity to get there, I think that's good and you just have to look: What has the highest effect for the house afterwards? Where do we find the best durability in which means? And maybe you have to do it completely individually per customer or per cohort. I don't know that yet either, but in the end, the fact that we're getting there, making that possible, that's one thing. And I think that's also important. And the other is: How do I implement it in reality?
Benjamin Kolb [00:35:35]:
That sounds like very sophisticated A/B testing in the end. If you break it down to individuals, of course you don't have a very large statistical size anymore, but if you think about it in terms of cohorts, of course you do. Even with the numbers that you have now - do you think essentially that that statistical size of the data is enough to ultimately do effective A/B testing? And that's always a bit of a question, or the amount of data always plays a central role in all of these issues.
André Lux [00:36:03]: Is it resilient? Is it not resilient?
Benjamin Kolb [00:36:04]:
Is that resilient? Exactly.
André Lux [00:36:05]:
Absolutely, that's relevant, and I think with the small-particle topics that are involved, if we're talking about mini-cohorts, then it's going to be difficult. So we do try to do A/B testing within the segments as well, so testing affinities against each other or testing five against three articles. What works better there? We do something like that, but basically we use A/B testing wherever we can use it at all, because it's not that easy. So we work with Kameleoon. I hope I'm not doing any advertising or anything. And we simply look at: "Which placements do we actually have available where we can run A/B testing, where we can design it effectively?", because the web devs are always involved to some extent in setting this up. And we also tested the placements, that is, which placements do we actually have available where we can A/B test the whole thing? So something like footer, header, checkout and so on and so forth. And that has to be set up first. And of course it's important to us that we have as many placements as possible where we can test it. And when we have it, then the advertising media are tested. So is signal color better than native? And so on and so forth. Then, of course, we do the whole thing with the effectiveness measurement, but at the moment we're doing it on the basis of case numbers, which are very high, and not in very small cohorts.
Benjamin Kolb [00:37:29]:
Question. Maybe one more step back to the cohorts. Are they fixed for you, that is, you think about them, model them and then use them for A/B testing or other purposes, evaluations, so to speak? Or do you plan to make it dynamic as well? I'm asking because we're currently doing a research project on this topic ourselves, where we're actually trying to use machine learning methods to automatically form such selective cohorts, which are more difficult to understand, but which might deliver more selective results. What is your idea?
André Lux [00:38:00]:
To be honest: We're still relatively far away from that, because we have so many construction sites that we're working on at the moment, and the resources don't allow us to push everything further at the moment and to take it to the extreme, so to speak. It would be nice if we could become more pointed or, I would say, more technical in this environment. But I think if we break up the segments, cohorts that we have even further, double them, then that would be good, because then we would be more finely divided and it would fit even better. But the future prospects that you have just described are still relatively far away.
Christian Kallenberg [00:38:37]:
Then here's a question from me that may not be quite as deep-dive as it once was, but perhaps a few steps back: What's your traffic situation like? How far along are you in terms of distribution via mobile app or web? My preconception about your readership is that they come much more via the web, or much more via desktop than via mobile device, than is the case with other publishers.
André Lux [00:39:01]:
So the prejudice, I would support that. That's totally true, because we realize quite simply that if we're within working hours, so to speak, then we're used desktop. If we start just before working hours, so, I don't know, between 6 and 8 or afterwards between 5 and 8 p.m., then it becomes more mobile again. So that's really mobile. Now during the day, these are just colleagues who use it heavily at work, access it heavily. So we are still very much desktop-based, but that is changing. We can clearly see that it is changing. We can see that a change is slowly taking place, but for us, the desktop is actually still very strongly represented, yes.
Benjamin Kolb [00:39:46]:
You mentioned earlier that you're also dealing with churn right now - subscribers who let their subscription lapse or cancel. What mechanisms do you currently have to counteract this? That is, perhaps early warning systems that tell you: "He might cancel his subscription soon," or even when it's already too late and he has canceled his subscription, so that you can perhaps win him back again. What do you have in place there in the area of data acquisition through to marketing, which will probably then start up?
André Lux [00:40:21]: That goes in the direction for me so automation. So which triggers can we set when and how? And what insights do we have about the customers? And when we talk about prevention, it's actually like this: We're not set up there yet at heise+. We actually have mechanisms. If someone is in the process of cancelling their contract, we make them another offer: "Don't you want to stay? You can get it for another 25 percent cheaper," and then we hope that they'll say, "Okay, yes, fine, I'll just test it again, or I'll just stay with it." That's one of those options. If someone actually quits, then we automate the process, i.e., we initiate a chain of contacts by e-mail, and then we send them a subsequent offer to return. So we do that in an automated way. That works something like the onboarding track, which is also not yet individual, but is significantly better also with reference to: "Download the app. Use the app," because we could then also send push messages or draw attention to new content. But things like that, which we use, for example, in the classic c't, where we see: "Okay, we have durability clusters," we see: "Okay, someone has been with us for one year, then has been with us for two years, three years, five years, ten years, over ten years, and so on." So we always look at how, so to speak, the transition from cluster 1, so shelf life span one year to two years, then into two years to three years, and so on and so forth. And that's where we identify warning signals at c't. And these warning signals say: "Actually, the transfer in relation to the previous year or the year before or whatever is much too low. And that's where we have to go in now. We have far too many departures within the cluster." Then we just have to look: What kind of approach is the right one? How do we do that? What exactly is the problem? And the nice thing here at heise+ is - well, with my c't, with the classic ones, it's always: You have to write to them somehow by e-mail or whatever and say: "Here, pay attention, c't is great and you have these added values and you also have a Plus subscription and it's really great and just use it more." So whatever triggers you can set there to make them say, "Okay, yes, it's great and I'll keep on the ball," and that's how you transfer them to the next shelf life cluster. In the case of heise+, we simply look - but we don't yet do this automatically and there is no trigger at all: What is the actual transition from cluster to cluster in terms of shelf life? Whereby we are actually first in the total shelf life: What is the shelf life at all in our case? We once had a year or so - it was about seven months - that we said: The average shelf life for heise+ is seven months, because a relatively large number go out at the beginning. And we have to work on that. But these shelf lives - a little digression - have now improved considerably because we offer three-month subscriptions, four-month subscriptions, or annual subscriptions, because then they're committed for the year, and of course we've significantly, significantly improved the shelf life. What's exciting for us is the phase when the first year or the three months expire, so to speak: How do customers behave afterwards? But I'll come back to the indicators. Of course, what we do and measure is: How many visitors do we have to a user? How often does he log in? What are his reading times? And these are the topics - we don't have that yet, I'll be honest with you. We're working on that. What are the signals that tell us: "Okay, someone who came five times before and read for three minutes, why is he now coming only once and reading for only two minutes? What is that? And how do we need to trigger him? Which device do we have to use to trigger him? And what content, if any, do we need to feed him to get him to come back more often?" Because everyone knows that in the digital context, the more often someone uses the digital product, the better the shelf life. We simply have to act on that. And these parameters that I just mentioned, we have to automate them, so to speak, so that impulses happen. We don't have that yet.
Benjamin Kolb [00:44:24]:
Yes, classically, you would then probably also be able to work retroactively with the data. Because you could also examine the subscriber dropouts, how the usage sequence was over the last three to six months, and then use the curves to recognize whether future users are at risk or not.
André Lux [00:44:41]:
Exactly, so via the raw data print we have exactly everything here in house that we can work on.
Benjamin Kolb [00:44:46]:
But that's good news.
Christian Kallenberg [00:44:47]:
And also pleasing - to find the transition - is the news that recently went through the media that Heise Medien has taken over t3n. t3n also has what I'll call a Plus model, similar to heise+. I think it's called "Pro-Membership". Can you say anything about the extent to which there is cooperation or best practice? Can you learn from them, or are you all one now, or what's the next step?
André Lux [00:45:14]:
So first of all great, right? I'm really looking forward to it.
Christian Kallenberg [00:45:17]:
André Lux [00:45:17]:
I think that's great for Hanover as a location, and I believe that we can mutually benefit each other, to put it succinctly. Basically, the DNA of t3n should remain exactly as it is. We're not going to interfere with it at all. They will continue to do their thing as they have done it so far, and they will continue to do it just as successfully. I think the worst thing would be for us to think that we know something better than they do. They should just continue to do exactly what they've been doing. And I think they're doing that very successfully. The first thing I thought when Mr. Heise communicated that here yesterday: I suddenly felt seven years younger. So digital pioneers, great. So all of a sudden I was a young target group again. What I think is great is that it complements this overall portfolio very well, so to speak, because we are not so well represented in these startup environments - in terms of what we are doing here at the moment. This is simply an addition to the overall portfolio. They remain self-sufficient, as I said, they continue to do their things the way they've been doing it. But of course - but that hasn't happened yet - there will be an exchange in various areas, simply to learn from each other and to see: What can I take over from us? They for themselves: What can they take over from us? But each in his own responsibility. Everyone makes their own decision.
Christian Kallenberg [00:46:39]:
André, we've been talking about subscriptions for so long now, and about subscription durability. What kind of subscriptions do you have and for how long?
André Lux [00:46:44]:
Well, I have an Amazon Prime subscription, I used to have a RoadBIKE subscription, but I haven't had it for a long time. So that's a bike magazine from Auto Motor & Sport - from Motor Presse, sorry, but they also do Auto Motor & Sport. Exactly, I otherwise have a telecommunications subscription, but since I have all the interesting magazines here in the house somehow, I don't really need to have a subscription in that context.
Christian Kallenberg [00:47:07]:
Benni and I had speculated that you might have a classic rock subscription. We've heard that you've had a musical career, from hard rock to jazz, back to hard rock. And so we had thought, "That André, he's a print reader of Classic Rock."
André Lux [00:47:23]:
Christian Kallenberg [00:47:25]:
André Lux [00:47:26]: So guitar and bass or something like that would actually be something, but unfortunately I'm just as bad as many others: I have YouTube there. The things that I want to learn on the guitar, I learn them all on YouTube. So that's the best thing there is to get better. And I actually used to have a lot of tablatures and booklets. But that didn't get me anywhere. So I'm three times faster through YouTube tutorials and that's also really fun because you get a sense of accomplishment so quickly. But, yes, guitar and music, that's a subject that's been with me for a long time. For a really long time. Cycling is a topic that has accompanied me since the beginning of my student days. And these are the things that I actually still do. Before Corona, there was also a lot of sport. I was out and about a lot in the gym. But that has decreased significantly now. But the consumption of red wine is high. No, it's not.
Benjamin Kolb [00:48:24]:
So the subscription is in the gym.
André Lux [00:48:26]:
Yes, it rests, exactly.
Christian Kallenberg [00:48:28]:
Okay, great. André, thank you very much.
André Lux [00:48:29]:
Christian Kallenberg [00:48:30]:
It was fun and, above all, insanely informative.
Benjamin Kolb [00:48:33]:
Yes, very interesting, really, I must say.
Christian Kallenberg [00:48:34]:
One of our longest deep dives we have done so far. Thank you very much.
André Lux [00:48:38]:
Christian Kallenberg [00:48:39]:
And we are keeping our fingers crossed that the 100,000 will be a reality by 2025 at the latest.
André Lux [00:48:44]:
We'll just talk again. Thank you.
Christian Kallenberg [00:48:46]:
André Lux [00:48:47]: