# LinkedIn as a self-experiment: 12 principles, real numbers > URL: https://4eck-media.de/en/blog/linkedin-as-a-self-experiment-12-principles-real-numbers/ > Language: en > Description: As of today, 26 July 2026, my LinkedIn profile has around 400 followers. For comparison: our TutKit.com YouTube channel has more than 24,000 followers, and on Facebook there are m… --- As of today, 26 July 2026, my LinkedIn profile has around 400 followers. For comparison: our TutKit.com YouTube channel has more than 24,000 followers, and on Facebook there are more than 110,000 spread across our own channels. We have run an agency since 2009, delivered more than 350 web and SEO projects and advise clients on AI and visibility. Yet on the one platform where our actual target group sits (managing directors and marketing leads in mid-sized companies), I have been all but invisible. It took me a long time to notice, because the other channels were performing and business is generally going well. As an agency, we are growing. It is still a little embarrassing. So we are starting an experiment. In public, with real numbers, including the months in which nothing works. I am about to join the linkedin game. So let’s play. ![Matthias Petri starts the LinkedIn self-experiment with around 400 followers](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-selbstversuch.avif "LinkedIn self-experiment: 12 principles and real numbers") ## The experimental setup Over the past few weeks I have worked through 13 issues of the Content Playbook, an English-language newsletter in which people like Lara Acosta and Jake Ward lay out their LinkedIn strategies. Both have built seven-figure businesses with LinkedIn as their main channel. From those issues I distilled 12 principles that keep coming up. Whether these principles work in a German B2B environment, I do not know. That is exactly what we are now putting to the test. The rules: One post a week for 12 months; I want to keep up that frequency at the very least until the end of July 2027. Once a month I evaluate the numbers and publish them here on the blog. We measure what counts: followers from the target group, conversations in direct messages, project enquiries and invitations to speak or appear on podcasts. I note impressions as well, but they are not the goal (I explain why in principle 9). What we are not doing: no engagement pods, no bought followers, no artificial scarcity with countdowns, no outrage posts for reach. Tactics like these do appear in the playbook. They suit neither us nor German mid-sized business. ## The 12 principles ### 1. The idea decides, the wording only amplifies A strong observation with mediocre wording can work. A weak idea with perfect wording stays on the ground. It sounds obvious, but it changes the workflow completely: most of the time now goes into the question of which real event from the past two weeks can carry a post. Client questions, objections raised in proposals, findings from audits. The wording comes afterwards. ### 2. The hook is 80 per cent of the post The first two lines decide whether anyone clicks “see more”. So the opening gets roughly half the writing time. It is written in reverse order: the content first, then the opening, and at the end a question that opens a conversation. By now I have built my own library of hook patterns, from the counter-intuitive claim to the sober sentence with no emotive words at all (as a former police officer, I call that one the police-report hook internally). So that design not only looks good but works measurably, our [web design agency](https://4eck-media.de/en/competencies/ui-ux-web-development-agency-4eck-media/) brings together UX, content and development. ### 3. Concrete moments instead of abstractions “Your website is not AI-readable” moves nobody. This sentence does: ask ChatGPT for the best providers in your industry and your region. If your competitor is named and you are not, you have a problem that your Google ranking will never show you. Pain is felt in moments, not in concepts. Every post needs a moment like that. For comparable tasks, our [SEO agency](https://4eck-media.de/en/competencies/seo-agency-4eck-media-we-will-take-you-to-the-top/) develops a sustainable search strategy from analysis, priorities and implementation. ### 4. Evidence beats assertions LinkedIn is full of invented success stories, and readers know it. Anyone who shows screenshots, real numbers and client voices wins immediately. For us that also means: we are our own case study. This experiment is principle 4 in action, with all the risk that entails. If the numbers are bad, they will still be here. ### 5. Opinion makes the expert, information is a commodity Now that anyone can generate advice articles with AI in seconds, pure information has become worthless. What remains is a point of view. So I am going to take a position, even when it gets uncomfortable: on website-builder promises, on AI scaremongering, on SEO consulting with no data behind it. Disagreement is expected and welcome. ### 6. Repetition beats originality The feed shows each post to only a fraction of your followers, and a week later everyone has forgotten it. That is why the most successful authors repeat 10 to 15 core messages in ever-new packaging. It goes against my gut, which constantly wants to produce something new. All the better as a thing to test: we define our core messages and, after 90 days, we may well bring a post that worked back out with a new opening. ### 7. Write for the client, not for colleagues The classic agency mistake: content that impresses other agencies. Write about agency topics and you attract agencies. Our posts are aimed at managing directors and marketing leads, and every text has to pass the mum test: would someone with no technical background understand what this is about? “llms.txt and schema markup” then becomes “we make your website as readable for AI systems as a book with a table of contents”. We cannot avoid technical terms entirely. But they get explained. ### 8. People buy identity and feeling, not a service Nobody buys a website project. What is bought is the outcome behind it: being found, being ahead of the competition, no longer having to worry about the technology. The posts should make exactly that picture tangible, with stories instead of lists of services. ### 9. LinkedIn is an access platform, not a reach platform The principle that convinced me most. There are millions of decision-makers on LinkedIn, and one of the playbook authors reports a post with 75 likes that brought in more than 200 enquiries. Reach is pleasant, but the value lies in access to the right 50 people. That is why we measure conversations and enquiries, not impressions alone. And that is why a follower count of 400 is less dramatic than it sounds. ### 10. A system beats a heroic effort Consistency comes from routines, not from motivation. Mine looks like this: content time in the morning before the first email (I am an early riser anyway, and the coffee is black), an idea store for every client question and every objection, and a stock of prepared posts for busy weeks. Larger pieces of content such as talks are mined for everything they have: one speaker deck supplies material for weeks. ### 11. Your own analytics are market research Once a month I look at which posts performed one and a half times better than my average. Those topics are validated and get built out, into articles, talks or series. One thing matters here: the numbers judge the posts, not me. Fallow phases are factored in; every author has them. ## Create effective content These articles show how strong ideas, clear messages and consistent distribution generate attention and contribute measurably to business goals. [![High-Attention Outreach: More Inquiries Than via Email (Case Study + Interview)](https://4eck-media.de/wp-content/uploads/2026/04/high-attention-outreach-4eck.avif "High-Attention Outreach: More Inquiries Than via Email (Case Study + Interview)")](https://4eck-media.de/en/blog/high-attention-outreach-more-inquiries-than-via-email-case-study-interview/ "High-Attention Outreach: More Inquiries Than via Email (Case Study + Interview)") [High-Attention Outreach: More Inquiries Than via Email (Case Study + Interview)](https://4eck-media.de/en/blog/high-attention-outreach-more-inquiries-than-via-email-case-study-interview/) Matthias Petri [![Renewable Energy Website with SEO](https://4eck-media.de/wp-content/uploads/2025/10/windpark-website-seo-720x480.avif "Renewable Energy Website with SEO")](https://4eck-media.de/en/blog/modern-website-with-a-strategic-content-base-our-project-for-leko-energy-gmbh/ "Modern website with a strategic content base: Our project for LEKO ENERGY GmbH.") [Modern website with a strategic content base: Our project for LEKO ENERGY GmbH.](https://4eck-media.de/en/blog/modern-website-with-a-strategic-content-base-our-project-for-leko-energy-gmbh/) Matthias Petri [![Bus layout for the Care Campaign by MeckCura](https://4eck-media.de/wp-content/uploads/2025/10/buslayout-gepflegt-kampagne-720x480.avif "Bus layout for the Care Campaign by MeckCura")](https://4eck-media.de/en/blog/received-a-well-deserved-punch-in-the-face-feedback-on-our-campaign/ "Received a well-deserved punch in the face: Feedback on our campaign") [Received a well-deserved punch in the face: Feedback on our campaign](https://4eck-media.de/en/blog/received-a-well-deserved-punch-in-the-face-feedback-on-our-campaign/) Matthias Petri [![Video tutorials for online printing.](https://4eck-media.de/wp-content/uploads/2025/10/video-tutorials-fuer-druckerei-viaprinto-1-720x480.avif "Video tutorials for online printing services")](https://4eck-media.de/en/blog/corporate-publishing-video-tutorials-for-the-professional-printing-house-viaprinto-de/ "Corporate Publishing: Video Tutorials for the Professional Printing House viaprinto.de") [Corporate Publishing: Video Tutorials for the Professional Printing House viaprinto.de](https://4eck-media.de/en/blog/corporate-publishing-video-tutorials-for-the-professional-printing-house-viaprinto-de/) Matthias Petri ### 12. Give first, then ask deliberately At most every fifth to tenth post is allowed to sell something. The rest delivers: knowledge, insights, mistakes others can learn from. And once a month there is a direct request, for an introduction or a speaker slot, for example. Apparently almost nobody does this, even though people like to help. We are testing that too. ![LinkedIn-Profil von Matthias Petri mit 403 Followern zum Start des Selbstversuchs](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-experiment.avif "LinkedIn-Experiment von Matthias Petri: Start mit 403 Followern") ## The benchmark Starting point on 26 July 2026: around 400 followers, zero live enquiries via LinkedIn. Interim goal by the end of January 2027: 2,000 followers from the target group, seven qualified project enquiries, at least two speaker or podcast requests. Every month an update appears here with the real numbers, the best and the worst posts and what we have adjusted. ## Logbook This section grows as we go. Here I document what has happened since the start. ### Week 1 (20 to 26 July 2026) On the profile: follow button activated, profile URL shortened to linkedin.com/in/matthias-petri-4eck, about text rewritten. Plus seven posts published: - “Your preview image has about one second”: why AI-generated poster infographics sink in the feed, and what a motorway billboard has to do with it (88 impressions, freshly online). - Website launch for Bolzakademie Rostock: holiday camps bookable directly, accessible to WCAG 2.2, told from a father’s perspective (330 impressions). - A networking confession: why I dodge events, and what a single meeting with AI consultant Katrin Berndt set off regardless (564 impressions, 5 comments). - A find during the relaunch for AWO Müritz: an emoji in the URL slug. It works technically; the question is whether you should (627 impressions). - The SEO tip you are almost ashamed of: the Yellow Pages. With Sistrix data on why business directories are having a comeback right now (244 impressions). - “I force my son to work in the agency unpaid”: a holiday-job argument as a parenting question, the strongest post so far (2,472 impressions, 16 comments). - Quiz question: what does the domain ending .eus stand for? Answered via the launch of ime.eus for a Tokyo-based corporate group (891 impressions, 3 comments). ![LinkedIn-Post von Matthias Petri mit seinem Sohn bei 4eck Media und 2.472 Impressionen](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-ferienarbeit-post.avif "Ferienarbeit bei 4eck Media: LinkedIn-Post mit 2.472 Impressionen") First observation after a week: the most personal post with the most provocative opening tops the list, with almost four times the second best. The purely technical posts trail behind. That matches principles 3 and 8, and it confirms the playbook thesis that the hook decides everything. ### Week 2 (27 July to 2 August 2026) Besides the posting, something has happened to the toolbox: I have started working highlights with Unicode characters into the posts, as structural anchors for subheadings and key statements. Because the usual generators for this are stuffed with adverts, our team went straight ahead and built its own tool: the [LinkedIn Text Formatter](https://4eck-media.de/en/linkedin-text-formatter/). Free, no sign-up, runs entirely in the browser. Using it consciously is part of the deal: screen readers struggle with these characters, so I use them sparingly and never for entire paragraphs. Seven posts, this time without an upward outlier. All figures as at 3 August: - Heiko, a headhunter from Berlin, now only signs software subscriptions with the shortest possible term: instead of renewing a subscription, people now build the cookie banner plugin themselves, with Grok and Kimi (302 impressions, 6 reactions). - “Does the idiot next door earn more than you? The answer costs 1 euro”: DIE ZEIT and its lesson in programmatic SEO with salary pages for every town from Malchow to Neubrandenburg (222 impressions, 10 reactions, 1 comment). - Four coding AIs, one briefing, 11 to 45 minutes: Grok 4.5 Heavy, Kimi K3 Max, ChatGPT Sol Ultra and Claude Opus 5 tested with the same prompt, as a document carousel (190 impressions, 1 comment). - The UX fail of the week: “Learn more” six times on a hotel homepage, found during the competitor analysis for a client project (114 impressions, weakest post of the week). - The counterpoint: why HelloFresh uses “Learn more” quite deliberately as a soft CTA, and why a slightly bad taste remains nonetheless (141 impressions, 2 comments). - Wildpark-MV.de: a complete relaunch from TYPO3 to WordPress, told from the perspective of the father who used to stand in front of the enclosures with his own children (253 impressions, 14 interactions, best interaction rate of the week). - “Search for our agency on Clutch. You probably won’t find us”: Claude found and recommended us anyway, and it turned into a job from Canada (226 impressions, 3 comments). The key figures after two weeks: the profile stands at 435 followers, we started at 403. The 7-day analytics show 2,869 impressions (up 50 per cent on the previous week), 1,187 members reached, 51 per cent of them outside my own network, plus 72 reactions and 13 comments. ![Wildpark-Beitrag auf LinkedIn](https://4eck-media.de/wp-content/uploads/2026/07/wildpark-linkedin.avif "Wildpark-Beitrag auf LinkedIn") **Three observations from week 2:** - Firstly, the outlier was missing. In week 1 the holiday-job post outshone everything with 2,472 impressions; this week all posts sat between 114 and 302. In return, the base carries: the network post from week 1 picked up another 700-plus impressions in the same period. Reach evidently also comes from the back catalogue, not only from what is new. - Secondly, the pattern from week 1 is confirmed: posts with a personal connection (the wildlife park as a family memory, the Canada job as a story) beat the purely specialist posts. The most sober post of the week, the hotel UX fail, performed worst. The interesting thing: precisely this post started a series. The HelloFresh counterpoint picked it up, and a third post on branding in the AI age is to follow. Whether a series makes up for weak individual figures we will see in the monthly update (principle 6, repetition beats originality). - Thirdly, the honest look at the demographics, and it packs a punch: 29 per cent of the members I reach are entry-level professionals, only 5 per cent managing directors. Measured against the target group (decision-makers in medium-sized businesses), I am currently reaching the wrong people. That is the price of broad topics and a work order for principle 7: more posts that deal with client problems rather than industry topics. [ ### Week 3 (3 to 9 August 2026) Seven new posts, plus measurable spillover from the previous weeks for the first time. The figures for the new posts: - Award nominations as a business model: why we regularly receive post in which only a four-figure invoice stands between nomination and award (556 impressions, 12 interactions, best post of the week). - The LinkedIn Text Formatter that our team built because the usual Unicode generators are stuffed with adverts (553 impressions, 10 interactions). - Client relationships in the AI era, told through a fishing trip to the Lofoten Islands with two clients (463 impressions, 18 interactions, best interaction rate of the week). - The “I-want-one-too effect” after initial meetings: what new clients regularly say when they have our minutes in their inbox 15 minutes later (180 impressions). - Search Console can now do social: a first evaluation of the TutKit YouTube channel, with the most-searched term being, of all things, “nude photos” (112 impressions). - WPML stays, Claude now does the texts: our translation workflow for multilingual WordPress projects (80 impressions). - The English language version as a GEO lever, published on the last day of the week and therefore still without a reliable figure. In the same period, further impressions came in from the previous weeks, including for the wildlife park relaunch (128), the Clutch and Canada post (101) and the branding post (84). The key figures after three weeks: 448 followers, we started at 403. In the seven days, 2,659 impressions, 11 per cent fewer than in the previous week, plus 814 members reached, 71 reactions and 6 comments. ![Auszeichnungen als Geschäftsmodell](https://4eck-media.de/wp-content/uploads/2026/07/ausgezeichnet-linkedin.avif "Auszeichnungen als Geschäftsmodell") **Three observations from week 3:** - Firstly, reach has dipped slightly even though the peak values have risen. In week 2 the best post stood at 302 impressions, in week 3 at 556. Nevertheless, the bottom line is a minus of 11 per cent, because the weaker posts fell off much further this time. One pattern I recognise after three weeks: posts that criticise a behaviour or give away a tool perform better than posts that explain something. - Secondly, the most important warning figure of the week: the share of reach outside my own network has fallen from 51 to 40 per cent. So I am increasingly reaching the people who already follow me and breaking into other people’s feeds less often. At 448 followers that is exactly the wrong direction, because growth happens outside your own bubble. Comments have also dropped from 13 to 6, with an almost identical number of reactions. Two explanations suggest themselves: several posts this week ended on a thesis rather than a question, and three of the seven posts revolved around specialist topics with a narrow audience. - Thirdly, the target group is shifting slowly. The share of entry-level professionals among the members reached fell from 29 to 24 per cent, while smaller companies with 2 to 10 employees moved up to 16 per cent. That is not yet the target group from the experimental setup, but it is heading in the right direction. For week 4 I am therefore setting myself two things: more posts that deal with concrete client problems, and consistently ending every post with a question again. Along the way, a tool emerged this week that grew out of the experiment itself: because I had started setting highlights with Unicode characters, and the available generators are stuffed with adverts, our team built its own [LinkedIn Text Formatter](https://4eck-media.de/en/linkedin-text-formatter/). Free, no sign-up, everything runs in the browser. Because screen readers struggle with these characters, I use them sparingly and never for entire paragraphs. There was also a concrete enquiry from a zoo here in M-V, because the marketing manager had noticed that we carried out the Wildpark-MV relaunch. It is not yet entirely clear whether the trigger came via the LinkedIn post or by other routes, such as a Google search. I am still finding that out. ### Week 4 (10 to 16 August 2026) The best week of the experiment, and of all things it is the post with the least professional relevance that tops the list. - A product sample from Japan that arrived at the agency after a wrangle with customs, for a client whose corporate identity and website we are currently developing (1,962 impressions, 12 interactions, by far the strongest post of the week). - “Spot the mistake”: a client website disappears from the Google index even though it runs perfectly in the browser (521 impressions, 13 interactions). - Under the radar, episode 1: the industrial company IVS Isolierung from Malchin, presented as an eight-page carousel (414 impressions, 21 interactions, best interaction rate of the week). - The UX fail in the Nordkurier newsletter, where the citizens’ ombudsman had his head cropped off in the teaser image (207 impressions). - 18,000 hidden casino spam pages and more than a dozen unknown administrators: two hacked WordPress installations from this month (204 impressions, only online for a few hours). - Five things about Slack that annoy me every day, and why we are bringing communication into our own tool Planora (188 impressions). - Four times 100 in the PageSpeed test for a WordPress site that has been running for two years (119 impressions). Further impressions came in from the previous weeks, including for the GEO post on the English language version (152) and the award post (148). The key figures after four weeks: 453 followers, we started at 403. In the seven days, 4,227 impressions, up 53 per cent on the previous week, plus 1,745 members reached, 62 reactions and 20 comments. **Three observations from week 4:** Firstly, the decline in reach from week 3 has reversed. The share outside my own network rose from 40 to 61 per cent, the members reached from 814 to 1,745. So I am breaking into other people’s feeds again, and that is mainly down to a single post: the Japan parcel alone brought almost as many impressions as the entire previous week. It was the most personal and least specialist post of the week, which confirms the pattern from week 1. ![Japan-Beitrag auf LinkedIn](https://4eck-media.de/wp-content/uploads/2026/07/japan-beitrag-woche-4.png "Japan-Beitrag auf LinkedIn") Secondly, comments have risen from 6 to 20. This week I closed with a question several times. Perhaps it helped. Thirdly, the demographics, and here the desired shift is finally happening: for the first time, the entry-level group is no longer at the top; instead, experienced professionals lead with 27 per cent. In terms of company size, firms with 51 to 200 employees now lead with 16 per cent, whereas in week 3 it was still micro-enterprises with 2 to 10 people. After four weeks, the target group reached is thus moving noticeably towards medium-sized businesses for the first time, which is where our clients are. What I take away for week 5: personal posts bring the reach, specialist posts bring the right target group. Both together in one week seems to work better than a week full of specialist topics. ### Week 5 (17 to 23 August 2026) Seven days, seven posts: - The hacked WordPress sites: 18,000 hidden casino spam subpages at one client, more than a dozen unknown admin accounts at a second (142 impressions, 4 reactions, 1 comment). - The relaunch for WMH Herion, a Bavarian specialist in drive technology, as a nine-page carousel (264 impressions, 7 interactions). - Mecklenburg-Vorpommern’s AI funding programme, aimed specifically at companies, consultants and coaches from M-V (354 impressions, 13 reactions, 2 comments). - The review of one month of daily posting, triggered by a printed and framed thought-provoking post by Björn Tantau (761 impressions, 16 reactions, 9 comments, the strongest engagement of the week). - A suggestion for how customers can be nudged towards more Google reviews with small give-aways such as Merci chocolates or stickers (215 impressions, 4 reactions, 1 comment). - The internal AI-first memo to our own team, with the question of which roles will no longer exist in their current form by 2028 (218 impressions, 4 reactions, 2 comments). - A five-page carousel with designed 404 pages from client projects (195 impressions, 4 reactions, 2 comments, 1 repost). Considerably more reach also came in from the previous weeks, above all from the Japan mousepad post from week 4: another 5,660 impressions in this window alone, plus 349 for the post about the vanished website. The key figures after five weeks: 466 followers, in week 4 it was 453. In the seven days, 8,538 impressions, up 76 per cent on the previous week, plus 4,765 members reached, 108 social interactions with 79 reactions, 18 comments, 1 repost, 5 saved and 5 forwarded posts. Three observations from week 5: Firstly, the daily writing routine is now genuinely a routine. The review of it was at the same time the post with the strongest engagement of the week, more comments and reactions than any specialist post. Secondly, a single older post carries the bulk of the growth. Of the week’s 8,538 impressions, roughly two thirds fall to the Japan mousepad post from week 4, which continues to circulate in other people’s feeds. The share outside my own network rose accordingly from 61 to 83 per cent. So the 76 per cent increase does not primarily describe seven strong new posts, but a single outlier with a long tail. Important for the context, so that no false pattern of success for the new content is derived from a single week. Thirdly, the pattern from week 4 is confirmed by the new posts themselves: the personal, reflective post about posting beats every specialist post when it comes to conversation, even when the specialist post, as with the M-V funding programme, deliberately addresses a target group. ![LinkedIn-Post Woche 5](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-woche-5.jpg "LinkedIn-Post Woche 5") What I take away for week 6: the first post with direct lead intent (M-V funding) is now live and can be tracked over several weeks, both in impressions and in actual enquiries. At the same time, it is worth watching whether, once the Japan post fades, the weekly impressions fall back to a lower but more stable level. ### Week 6 (24 to 30 August 2026) Second week in a row without a gap: seven new posts for seven days. - A chat excerpt on the Elementor security vulnerability, linked to our own stance against page builders: “Point-and-click was yesterday, prompt-and-go is today” (102 impressions, 3 reactions). - Why the agency is called “4eck” and how to pronounce the name correctly, the origin story from Viereck near Pasewalk (1,168 impressions, 33 reactions, 10 comments, the strongest post of the week). - A relaunch that began with a second opinion rather than a briefing: the tax consultancy SEDULUS from Cottbus (219 impressions, 5 reactions). - 200 Google reviews after more than ten years, 199 of them 5 stars (672 impressions, 11 reactions, 2 comments). - A deliberately bold business tip on positioning: simply pay invoices straight away (268 impressions, 11 reactions, 1 comment). - Our own mastermind group with executive coach René Wasmund and Hannes Schröder from Outness (936 impressions, 27 reactions, 1 comment, the second-strongest engagement of the week). - Bye-bye Slack, hello Planora Meet: our own team communication moved into the tool we developed ourselves (125 impressions, 6 reactions). ![Mastermind Linkedin](https://4eck-media.de/wp-content/uploads/2026/07/mastermind-linkedin.avif "Mastermind Linkedin") The key figures after six weeks: 477 followers, in week 5 it was 466. In the seven days, 4,359 impressions, down 60 per cent on the previous week, plus 1,675 members reached, 207 social interactions with 192 reactions, 14 comments and 1 post forwarded on LinkedIn. Three observations from week 6: Firstly, the 60 per cent drop is deceptive. This week’s seven new posts come to 3,432 impressions in total, more than the 2,149 of the new posts from week 5. The lower overall figure arises solely because the Japan mousepad post from week 4 is no longer having an effect. Without that one-off outlier the picture is this: the actual weekly content performance has risen, not fallen. Secondly, for the third time in a row, personal posts lead the field. The story behind the agency name and the reflection on our own mastermind group take first and second place for impressions, reactions and comments, well ahead of the specialist posts on SEDULUS or on positioning. All 14 comments of the week come exclusively from the new posts; not a single one arrived belatedly on older posts. Thirdly, “Rostock and surroundings” appears among the top locations for the first time (12 per cent), together with a cleaner distribution of company sizes: 11 to 50 employees instead of, as in week 5, predominantly large corporations with more than 10,000 staff. The timing fits the M-V-focused funding post from week 5 and could be a first indication that deliberately regional posts actually shift the target group reached. One week is too little for a reliable statement, however. And it included the first two leads that reached me by email, where the groundwork had been done through trust-building on LinkedIn. One was an introduction concerning the AI funding programme, the other a tracking audit. What I take away for week 7: comparing pure content performance (without viral outliers) is a more honest yardstick for the self-experiment than the raw total of impressions. And the trend towards personal, reflective posts is now too clear to dismiss as coincidence. ### Week 7 (31 August to 6 September 2026) Third week in a row without a gap: seven new posts for seven days. - Service-area maps from client projects as a carousel, a UX detail against the uncertainty before the first enquiry (291 impressions, 4 reactions). - The labelling requirement for AI images as a supposed coup by the stock photographers’ lobby, satirically sharpened (2,965 impressions, 7 reactions, 2 comments, by far the strongest post of the week). - A security headers check using the new client project PROMIX as an example (270 impressions, 1 reaction). - Why your own QR codes in corporate design are worthwhile, including a reference to the QR code generator we built ourselves (514 impressions, 9 reactions, 1 comment). - A punchy business tip: respond to enquiries extremely quickly, by phone if need be (739 impressions, 16 reactions, 8 comments, the strongest conversational engagement of the week). - A personal encounter with the local competitor Hannes Richter at the Müritz Connect summer party, and a plea for open exchange instead of competitive thinking (308 impressions, 7 reactions, 4 comments). - Two server migrations, the same effect in Search Console: faster response times, more crawling (99 impressions, 4 reactions). The key figures after seven weeks: 481 followers, in week 6 it was 477. In the seven days, 6,315 impressions, up 45 per cent on the previous week, plus 3,178 members reached, 117 social interactions with 95 reactions, 17 comments, 1 repost and 4 saved posts. ![LinkedIn-Beitrag zu Stockfotografen und KI-Bildkennzeichnung](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-stockfotografen.avif "LinkedIn-Beitrag zu Stockfotografen und KI-Bildkennzeichnung") **Three observations from week 7:** Firstly, this week also has its own outlier, but this time not through spillover from a previous week but through a fresh post: the satirical post on AI image labelling alone came to 2,965 impressions, almost half of the entire weekly total. The seven new posts together reach 5,186 impressions; the rest is spread across older posts that continue to have an effect. Secondly, this time it is not the most personal post that wins but the most pointed. Unlike in weeks 5 and 6, it was not a review or a reflection but a pointed, almost polemical thesis that performed best. In terms of engagement, by contrast, a practical tip (respond quickly) leads, having drawn the most comments of the week. This suggests that reach and willingness to converse depend on different types of post: sharpening brings impressions, a concrete tip with a position of your own brings discussion. Thirdly, despite the overall strong week, the follower count continues to grow only slowly: plus 4 compared with gains of 11 to 13 in the previous weeks. So high reach and many interactions do not automatically translate into new followers. An important pointer for the self-experiment: look at impressions and follower growth separately rather than equating them. This week I also commented little on other people’s posts. What was nice was that I was approached three times about my LinkedIn engagement in real life. The people from the region who are online on LinkedIn notice it, even if they do not like or comment. What I take away for week 8: it pays to distinguish more deliberately between pointed opinion posts (for reach) and concrete, actionable tips (for conversation), rather than mixing both in one post. ### Week 8 (7 to 13 September 2026) After two weeks of uninterrupted daily posting, this time a real gap: six new posts in seven days, with a break on 12 September. - The review of eleven years of looking after fischkaufhaus.de, the online shop of the Müritzfischer, including two Shop Usability Awards and the current Seafood Star 2026 (443 impressions, 15 reactions, by far the strongest post of the week). - The hack of Berlin’s state IT via a forged fake-captcha page, with reference to two of our own client projects where the same scam had already been seen live (222 impressions, 4 reactions). - The relaunch for furniRENT from Austria, outfitter for owner-run hotels, with a hotel refurbishment calculator and an interactive project map (231 impressions, 5 reactions, 2 comments, 1 repost). - A Facebook ad for the open day of a funeral home in Waren (Müritz): 50,000 impressions, 200 clicks, 0 registrations, and the question of whether commitment is declining in general (310 impressions, 5 reactions, 2 comments). - A defence of WordPress against blanket bashing: implementation quality rather than blaming the system, with reference to WordPress’s market share of around 59 per cent (210 impressions, 6 reactions, 1 comment). - Two websites at two different hosting providers that locked Google out via firewall and bot-protection rules (117 impressions, 4 reactions). - Shop system migration from PrestaShop to WooCommerce for Wald-zu-Tisch (289 impressions, 6 reactions) Reach continued to come in from the previous weeks, around 1,155 impressions in total. Surprisingly, the biggest share of it came not from the viral AI images post from week 7 (187 impressions in this window) but from the two-week-old mastermind post from week 6 with René Wasmund and Hannes Schröder (205 impressions). The key figures after eight weeks: 491 followers, in week 7 it was 481. In the seven days, 2,688 impressions, down 57 per cent on the previous week, plus 800 members reached, 57 social interactions with 52 reactions, 4 comments and 1 post forwarded on LinkedIn. ![LinkedIn FKH](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-fkh.avif "LinkedIn FKH") **Three observations from week 8:** - Firstly, the pattern from the previous weeks is confirmed once again: without a viral outlier of its own, total reach falls markedly, even though the seven new posts together, at 1,533 impressions, are well within the range of the previous weeks. The decline mainly affects the peak values, not the baseline. - Secondly, a small collaboration has come about with a fellow SEO professional, initiated by a helpful comment of mine on his posts. - Thirdly, this time it was not a pointed opinion post that performed most strongly but an eleven-year-old client story: the review of looking after fischkaufhaus.de. This suggests that even unspectacular topics such as continuity and long-standing client relationships go down well when they are backed up with concrete details, such as awards, figures and a personal connection. What I take away for week 9: the strongest performance this time came from an editorially rather unspectacular case study post, not from a pointed opinion post. In the weekly evaluation, it pays to look more deliberately at such quieter successes too, and not only at the loud outliers. ### Week 9 (14 to 20 September 2026) Seven new posts in seven days, because two posts appeared on 20 September. After the gap in week 8, that looks like a small catch-up effect. - Planora, our own tool that replaced Jira in April, now grown into a complete AI-first business suite (285 impressions, 8 reactions, 1 comment). - The online sick-note system for schools and nurseries of the town of Neustrelitz, more than 21,000 absence notifications in the past year, with a quote from Tobias Menke, Head of Schools and Sport (429 impressions, 16 reactions, 4 comments). - A line corrected in 2022 in the Google Business Profile of the Fischerhütte Sietow which, according to the Local Guide profile, has since generated more than 2.4 million views (658 impressions, 13 reactions, 2 comments). - A PageSpeed quiz: 99, 100, 100 and 100 points for a WordPress website by HRM CONSULTING GmbH, as a counterexample to blanket WordPress bashing (741 impressions, 14 reactions, 7 comments). - A two-month interim review of the self-experiment: 500 followers, two leads without a job, one small collaboration, and the observation that personal posts generate more conversation than specialist posts (2,508 impressions, 41 reactions, 19 comments). - llms.txt: 135 days of bot logs evaluated, 64 requests from 14 different bots, plus the finding that crawlers expect language-specific files (12,063 impressions, 53 reactions, 11 comments, by far the strongest post of the week and of the entire self-experiment). - An anecdote on image rights and negotiating power, starting from a price surcharge for client photos in a coworking space, with the Theodore Roosevelt example from Jack Nasher’s “Deal” (423 impressions, 8 reactions, 3 comments). The key figures after nine weeks: 537 followers, in week 8 it was 491. In the seven days, 18,136 impressions, up 575 per cent on the previous week, plus 284 social interactions with 176 reactions, 58 comments, 45 saved posts and 5 posts forwarded on LinkedIn. **Three observations from week 9:** Firstly, the first viral outlier of our own in the entire self-experiment: the llms.txt post alone comes to 12,063 impressions, almost two thirds of the entire weekly total, and thus clearly surpasses the previous record holder from week 7 (AI images post, 2,965 impressions). Remarkably, it is not a pointed opinion post but a data experiment with real figures collected over 135 days. ![LLMS.TXT-Beitrag](https://4eck-media.de/wp-content/uploads/2026/07/llms-linkedin-beitrag.avif "LLMS.TXT-Beitrag") Secondly, the follower gain of plus 46 (from 491 to 537) is the biggest jump in the entire self-experiment, more than three times the previous best from week 8 (plus 10). In terms of timing, this coincides with my own 500-follower interim review post, which talks openly about the successes and disappointments of the experiment. Honest self-reflection and the viral specialist post may have reinforced each other here. Thirdly, comparing the engagement figures gives a nuanced picture: the interim review drew the most responses of the week with 19 comments, even though the llms.txt post had more than six times as many impressions. So reach and willingness to converse continue to diverge, as already observed in week 7. What I take away for week 10: data-driven posts with real, self-collected figures beat everything else in week 9, more clearly than any opinion post before. It pays to evaluate your own raw data more often and turn it into a post, instead of only taking positions. ### Week 10 (21 to 27 September 2026) Seven new posts in seven days, a clean one-post-per-day rhythm without a gap. - An SEO post on star ratings in search results, using the example of two Karrierebibel articles with more than 6,000 votes each, even though there is no voting button to be found anywhere (531 impressions, 3 reactions, 1 comment). - A joint photo shoot with four consulting clients at the same time, namely Stefan Siewert, Nicole Hirscher, Katrin Berndt and Karsten Böhm, as an example of how an agency creates additional value in the AI era (1,111 impressions, 22 reactions, 4 comments). - A personal reflection on the film Idiocracy, prompted by a conversation with a client from Berlin (473 impressions, 7 reactions, 3 comments). - A UX critique of encrypted, desaturated Google Maps embeds on holiday websites, with an example from Carinthia (555 impressions, 4 reactions, 9 comments). - A practical case on IP reputation and email deliverability at T-Online after a server move, with a three-step guide (740 impressions, 10 reactions, 1 comment). - The direct follow-up to the viral llms.txt post from week 9: Markdown Negotiation as, in my view, the more important GEO lever, with figures on bot requests from ChatGPT-User, Meta and Applebot (4,970 impressions, 32 reactions, 17 comments, by far the strongest post of the week). - A case study on the precious metals shop of Goldmichi from Rostock, with live market prices and three parallel tax logics (100 impressions, 5 reactions, no comments). ![LinkedIn-Fotoshooting](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-fotoshooting.avif "LinkedIn-Fotoshooting") Reach continued to come in from the previous weeks, around 6,679 impressions in total. The original llms.txt post from week 9 alone brought another 3,940 impressions in this window, one week after publication. The key figures after ten weeks: 564 followers, in week 9 it was 537. In the seven days, 15,159 impressions, up 16 per cent on the previous week, plus 233 social interactions with 141 reactions, 46 comments, 0 reposts, 40 saved posts and 6 posts forwarded on LinkedIn. **Three observations from week 10:** Firstly, the llms.txt topic from week 9 holds the top position in week 10 as well: the direct follow-up on Markdown Negotiation comes to 4,970 impressions, which is well below the previous week’s viral outlier (12,063 impressions) but clearly above all other posts this week. A topic that catches on once evidently carries the follow-up post too. Secondly, reach is spread more broadly in week 10 than in week 9. The seven new posts bring 8,480 impressions in total, more than the 6,679 impressions from older posts. In week 9, conversely, a single viral post had carried almost two thirds of the total reach on its own. Without that one outlier, the weekly growth in impressions falls accordingly from plus 575 per cent to plus 16 per cent, with operations continuing but more evenly. Thirdly, the rhythm of seven posts in seven days continues. The follower gain remains noticeable at plus 27 (from 537 to 564), albeit smaller than the record jump from week 9 (plus 46). What I take away for week 11: a follow-up post on a successful topic can come close to the original if the topic has real discussion value. With strong topics, it pays to deliberately turn them into a small series rather than moving on after one post. ### Week 11 (28 September to 4 October 2026) Seven new posts in seven days, again exactly one per day. The Wörthersee post from 28 September belongs to this week; it was deliberately left out of the evaluation for week 10. - The Wörthersee as a co-branding element: because the television series “Ein Schloss am Wörthersee” evokes holiday memories for many guests from northern Germany, the lake appears on the map section for apartment houses in Villach and at the Faaker See, and on the map of the new campsite at the Jabeler See it is the Müritz (326 impressions, 8 reactions, 3 comments). - A call for a code of honour for success screenshots on LinkedIn with four rules: show the domain, measure PageSpeed on mobile, use current and complete reports, and internal data such as Search Console or revenue may remain anonymous (218 impressions, 8 reactions, 3 comments). - A practical tip for Claude users: in the usage overview there is a one-off free reset for the five-hour and weekly limits, redeemable until 22 October (394 impressions, 13 reactions, 6 comments, 1 repost). - The story of Timo Hannemann, who bought a knife in Japan because of the photo of a 70-year-old knife smith, as an example of authentic positioning, with the question to readers of how much they show who they are (754 impressions, 6 reactions, 2 comments). - Two spam protection methods for contact forms without a captcha service, a honeypot field and a signed time trap, expressly released for others to copy (1,402 impressions, 4 reactions, 2 comments). - The new “Agentic Resource Discovery” audit in PageSpeed Insights, which raised the agentic browsing score for our own website from 3 to 4 out of 4 audits passed (253 impressions, 5 reactions, no comments). - Concrete answers to guest questions on hotel websites, hung on the question of whether you can charge your e-tron on site: a car park for SUVs, a parking space with charging station, a lockable bike box and precise directions on a dedicated page of a client project, plus the tip to hosts to write down the most frequent guest questions and answer them on the website. Published on Sunday, the last day of the week (2,302 impressions, 7 reactions, no comments, by far the strongest post of the week). Reach continued to come in from the previous weeks, around 1,987 impressions in total. The largest share came from the two-month interim review from week 9 with 542 impressions, followed by the T-Online post from week 10 with 373 and the Markdown Negotiation post from week 10 with 274. The llms.txt post, which came to 12,063 impressions in week 9, brought fewer than 100 in this window. The guest questions post from Sunday will show up as spillover in week 12, because by the time of the count on Monday around 780 further impressions had already come in. The key figures after eleven weeks: 574 followers, in week 10 it was 564. In the seven days, 7,636 impressions, down 50 per cent on the previous week, plus 4,700 members reached, 79 per cent of them outside my own network, and 92 social interactions with 71 reactions, 18 comments, 1 repost, 2 saved posts and 0 posts forwarded on LinkedIn. ![Beitrag über Parken in Hotels](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-niederland.avif "Beitrag über Parken in Hotels") **Three observations from week 11:** Firstly, after two exceptional weeks, reach falls back to a more normal level. 7,636 impressions are just under half of the previous week (15,159), but still almost three times week 8 (2,688). Members reached fall from 10,269 in week 9 via 8,085 in week 10 to 4,700 now, social interactions from 208 to 92 (minus 56 per cent). Engagement is thus dropping somewhat more sharply than reach. Secondly, this time two practical posts carry almost half of the weekly reach. The guest questions post (2,302) and the spam protection post (1,402) come to 3,704 impressions together, which is 48.5 per cent of the weekly total. Overall, the seven new posts bring 5,649 impressions, i.e. 74 per cent of the weekly reach. In week 10 it was 8,480 out of 15,159, i.e. 56 per cent; the spillover from the llms.txt posts has evidently run its course. Thirdly, reach and conversation diverge again. The two strongest posts drew only 2 comments between them. The most responses, with 6 comments, went to the Claude tip (394 impressions), followed by the Wörthersee and the code of honour posts with 3 each. What I take away for week 12: how-to guides and concrete client examples carry the reach. The two strongest posts, both guides, ended without a question to the readers and came to 2 comments between them on 3,704 impressions. In week 12 it is worth testing whether an explicit closing question under a guide changes that. ### Week 12 (5 to 11 October 2026) Seven new posts in seven days, making it the third time in a row with one post per day and no gap. - Monday 5 October: the free LinkedIn formatter that converts text into bold, italic, capitals and fifteen Unicode fonts, plus the request to readers to say which three to five structural characters and emojis they are still missing (1,126 impressions, 5 reactions, 10 comments). - Tuesday 6 October: the renewed corporate identity for the Hoffnungsstiftung in Berlin-Pankow, with a new colour concept, reworked logo, style guide and everything from stationery to concert programme, finished in time for the foundation’s tenth anniversary, including a thank-you quote from the foundation’s board after the concert (307 impressions, 7 reactions, no comments). - Wednesday 7 October: the relaunch for Genuss-Ferien, the nature and beach campsite at the Jabelscher See, with seven language versions, its own overview map of the region and before-and-after screenshots for PageSpeed, accessibility, security and agent readiness (380 impressions, 13 reactions, no comments). - Thursday 8 October: a look behind the scenes at Lenz Wild in Woldegk, with game hanging in the cold store and the end product as game burgers and minced-game sticks, plus the guessing question of what the stag still weighed without head and innards (5,326 impressions, 35 reactions, 15 comments, by far the strongest post of the week). - Friday 9 October: why we take on three to four school work-experience students every year, even though it never pays off in business terms, and what they learn with us about the influence of AI on the profession (345 impressions, 14 reactions, no comments). - Saturday 10 October: rotating wind turbines on the project map for LEKO ENERGY from Potsdam, built in as an SVG animation so that the client can place new sites themselves later (132 impressions, 6 reactions, no comments). - Sunday 11 October: clients now brief us with ChatGPT, Claude and Gemini, the middleman between briefing and finished website is increasingly disappearing, hence the focus on more complex projects where someone puts their name to the result (247 impressions, 10 reactions, 4 comments). Reach continued to come in from the previous weeks, around 3,515 impressions in total. Of that, 2,841 fall to the guest questions post from 4 October, which thus gathered more impressions in this window than on its day of publication (2,302). Behind it come the spam protection post from 2 October with 138 and the ARD post from 3 October with 84 impressions. The key figures after twelve weeks: 584 followers, in week 11 it was 574. In the seven days, 11,378 impressions, up 49 per cent on the previous week, plus 6,737 members reached, 80 per cent of them outside my own network, and 126 social interactions with 95 reactions, 30 comments, 0 reposts, 1 saved post and 0 posts forwarded on LinkedIn. ![Wildbeitrag auf LinkedIn](https://4eck-media.de/wp-content/uploads/2026/07/linkedin-wild.avif "Wildbeitrag auf LinkedIn") **Three observations from week 12:** Firstly, reach recovers after the previous week’s slump, carried by a single post. The game post, with 5,326 impressions, accounts for 47 per cent of the weekly total and is thus, after the llms.txt post from week 9 (12,063), the second-strongest post in its publication window of the entire self-experiment, ahead of the Markdown Negotiation post from week 10 (4,970). Without it, impressions would stand at 6,052 and thus below the previous week’s figure. Compared with week 10 (15,159), 25 per cent are still missing; for members reached (8,085 in week 10) it is 17 per cent. Secondly, the assumption from week 11 about the closing question turns out mixed. All 29 comments under the new posts went to three of the five posts with a closing question (game 15, formatter 10, AI briefing 4); the two posts without a question stayed at zero. However, two posts with a question also received not a single comment (work-experience students, wind turbines). So a question seems to be necessary, but is not enough on its own. What stands out is that the two successful questions were easy and concrete to answer: guess a weight or name three to five characters. For the formatter post, the comment rate of 10 comments on 1,126 impressions is 0.9 per cent; for the game post, 0.3 per cent. That remains an assumption that can only be substantiated over several weeks. Thirdly, follower growth still does not keep pace with reach. At plus 10 (from 574 to 584), it is exactly on the previous week’s figure, even though impressions rose by 49 per cent and the strongest post was seen 80 per cent outside my own network. Compared with the plus 46 from week 9 and the plus 27 from week 10, this looks like a normalisation, entirely without viral outliers. What I take away for week 13: a personal look behind the scenes at a client, combined with a guessing question, brought the most reach and the most comments of the week. In week 13 it is worth deliberately placing a question with a simple, quickly answerable reply under a specialist or client post and seeing whether the comment pattern repeats. If you want to follow the experiment: on LinkedIn you will find me as [Matthias Petri](https://www.linkedin.com/in/matthias-petri-4eck/), where the self-experiment is running live. And if you are struggling with LinkedIn yourself right now, rest assured: the two of us were in the same boat only recently.