65 articles
Before you speak to an agency, there is one work sample in plain view that nobody has picked out for you: its own website. Five checks that take a quarter of an hour and need no judgement of taste: speed for real visitors, what loads before you consent, who stands behind the legal notice, whether the site works without a mouse and whether the references can be verified.
When three proposals promise the same thing, the one line that differs decides: the price. That is not a negotiation question but a brand question. How many clients see any difference between providers at all, why the number wins when they do not, what clients read into a price, what a brand is worth when it comes to price, and how to make your difference comparable.
Google and Apple change your opening hours when users report something and check the change against public sources, first of all your website. When your core details read differently in different places, it is not your profile that decides but what your own site says. How often customers run into wrong details, how a change gets into your profile, which places really count, what happens with holidays, a move or a new number, and how to set one source everyone believes.
The chatbot is not the problem. The second in which it runs out of answers is. Gartner surveyed more than 3,500 customers in February and March 2026: only seven per cent used a chatbot for their most recent service issue, and only 27 per cent would try one again after a bad experience. What happens at that moment decides the case - and whether anyone ever opens the tool again.
The answer is on your site, and the search box still says no results. The reason is not missing content but mechanics: the built-in search compares strings of characters, the visitor is looking for a meaning. What WordPress and Shopify actually do, which queries fail, who you lose at zero results, what our sample of ten websites showed, and how a search finds meaning instead of letters.
Translation is the cheapest part of a second language. Counting words alone, the initial translation accounts for roughly 8 per cent of the five-year total and maintenance for the rest. And a mediocre translation costs less than a second language Google cannot connect to the first: of the 20 ATX home pages, 13 carry any language annotation at all, four of them complete.
The welcome email is opened more often than any other message you will ever send - and unsubscribed from more often too. What the benchmarks actually show, what German and Austrian law require in email one, why the first hour carries almost half of all clicks, and which three jobs the first three emails do.
After the rebuild you want to know whether it is better. The numbers that would answer that are still there today, but they expire: Google Analytics deletes event data after 2 or 14 months depending on the setting, the field data history reaches back 40 weeks, Search Console 16 months. Proving a 30 per cent improvement on 20 enquiries a month takes about 23 months of measurement.
The group deciding on your proposal runs to between five and 16 people across as many as four functions, according to Gartner. In the same survey, content tailored to a single individual has a 59 per cent negative impact on that group reaching consensus. How to build a proposal that works without you in the room, how long it actually binds you, and what to do when nothing comes back.
Google answers the question in its own documentation, and the answer sits in brackets: "(No, it won't)". We measured when Google last fetched all 55 posts on this blog: a median of 28 days, a range of 5 to 62, and the three oldest posts were fetched sooner than the average. What an update actually triggers, and how to tell which page needs one.
The cookie banner is the visible part and the one with the smaller ceiling. An ordinary page loads a median of 79 third-party requests, 25 European supervisory authorities are checking transparency obligations in a coordinated action through 2026, and the legal basis for US tools rests on a decision now under appeal. The three points that matter, with the provisions you can look up.
A quote prices a project; from launch day onwards a website is an operation. 11,334 new security vulnerabilities in the WordPress ecosystem in 2025 alone, 91 per cent of them in add-ons, a weighted median of five hours to the first attack: the line items no quote can contain, with prices you can look up and a five-year calculation.
Three routes lead to a picture on the screen, but only one leads to a right in it. In February 2026 a Munich court held that generated logos belong to nobody, not even to the person who produced them. What a stock licence actually permits, why a missing credit doubles the damages, and which image belongs in which place.
Since 7 May 2026 Google has shown no FAQ rich result at all, for anyone. If you built the page for that one feature, you now have a page without a purpose. At the same time, half of all customer service journeys no longer start on your own website but on Google, YouTube or ChatGPT. What that means for your answers, and how to write them so they hold up where you do not control the layout.
The design quote is the smallest line on the bill. Britain's roads authority produced its new identity entirely in-house, at no external cost, and still spent more than the whole changeover had been budgeted for. What trade mark protection actually costs, what a domain change does to your visibility, and how to spot that a new name will not fix your problem.
83 per cent of people who are asked for a review go on to write one. Yet most businesses treat the ask like an awkward favour: too late, in the wrong channel, or with shortcuts that are outright banned. When asking is allowed, when it legally becomes advertising, and why your reply to a review matters more than the review itself.
In small accounts the ad almost always points at the homepage, because the homepage is already there. Yet 57 per cent of viewing time falls on the first screenful, and Google explicitly judges whether the destination meets the expectation your ad copy created. When a dedicated page earns its keep, when the service page you already have is enough, and from what volume you can test at all.
Every business says no far more often than yes. Yet there is a template for the yes and none for the no. 64 per cent of job seekers never get a final answer, and in the trades up to two thirds of enquiries go unanswered. What belongs in a good no, how fast it has to go out, and where the friendly sentence turns into a legal problem.
Connecting two tools is an afternoon's work these days. Whether it helps is decided before that: an integration moves data, it does not make decisions. If the rule about who does what and when is missing, the connection only speeds up the confusion. Here is how to tell the two cases apart in five minutes.
A classifier that files 95 out of 100 messages correctly sounds solid - until you look at the five it gets wrong. The two mistakes do not cost the same: a newsletter in the enquiry lane costs twenty seconds, a real enquiry in the bin costs the customer. Here is how to build the sorting so it only makes the cheap mistake.
Accessibility gets discussed as an obligation, which is exactly why so little happens: the obligation does not apply to most small providers. The reason to do it anyway is a different one. Captions, transcript and alt text are the only place where the content of sound and image exists as text.
Questions now get answered before anybody clicks. That does not devalue good pages, it shifts their job: they have to work as a source. What a citation is measurably worth, which queries are affected, why there is no markup for it, and how a citable page is built.
Twelve hours saved per week sounds like EUR 2,850 a month. It only reaches your accounts if it passes through one of three doors. What an hour of work really costs according to Eurostat, where the saved time actually goes according to Danish register data, and how to spot a calculation that will not hold.
Your sending tool reports 98 per cent delivered and still nothing happens. Delivered only means accepted, not read. What Google and Microsoft have actually required since 2024 and 2025, why a DMARC record enforces nothing on its own (our own measurement across 20 corporate domains), and what a 0.3 per cent complaint rate means in real clicks on a small list.
AI use at work rose from 59 to 75 per cent within a year, while the share of companies without a policy stayed at a third. Two thirds of AI access on corporate devices runs through personal accounts. What has been mandatory since February 2025, what providers actually do with your inputs, and what the rule looks like for a small team.
Gartner predicted that at least 30 per cent of GenAI projects would be abandoned after proof of concept. Reality overtook the forecast: in 2025, 42 per cent of companies reported scrapping the majority of their AI initiatives. The reasons are strikingly uniform - and all five patterns are visible before the budget is spent.
A calendar link does not automate scheduling, it hands the work to the person enquiring. A meeting has two rates: the booking and the attendance. What the solid evidence says about reminders is far more sober than the vendor promise.
The team page is where an enquiry is decided, and it usually shows a purchased smile. Since 2 August 2026 the picture is a legal question too, and Google's own rating guidelines list invented information about your team as the worst possible verdict. What belongs there instead.
In 2025 Google called off the cookie deprecation and then retired the replacement project. That removes the reason most companies built first-party data in the first place. The value is still there, but it sits somewhere else: in consent, in resolvability and in recency. Here is the calculation.
The share of AI citations coming from top 10 results fell from 76 to 38 per cent in seven months. At the same time, ChatGPT and Google no longer take the hard product facts from your page at all, but from a data channel. What that means for product and service pages.
With 20 responses your NPS swings by plus/minus 32 points. At that size the metric no longer measures anything you could steer by. What the score actually does, why its inventor revised it in 2021, and which two sentences in a survey genuinely carry weight.
95 per cent fail: the number is everywhere. It comes from an MIT report that contains it five times - counting something different each time. What hangs on that, where we quoted it wrong ourselves, and how to check a number in five minutes.
95 per cent of B2B marketers use AI, according to the Content Marketing Institute, and 87 per cent got more productive. Only 39 per cent saw their content perform better. Why that gap exists, and how to turn one researched piece into many formats without diluting it.
According to Recurly, poor onboarding is the single biggest driver of voluntary cancellations. What belongs in the first two weeks after signature, which steps you can automate and which ones you should keep personal.
On average, only 43 per cent of a planned price increase survives contact with the customer. The rest disappears in conversations nobody prepared for. How to build proposals that justify their price - and what to say when the discount question comes anyway.
41 per cent of companies now use AI, and a third of them pay more than they planned. The difference between an agent and a fixed workflow decides which side you end up on. Four tests that settle the question in ten minutes.
A language model does not read a brand manual. And it forgets rules measurably: once three constraints apply at the same time, accuracy drops by more than 40 per cent. What that means for your style guide, where machine-readable brand rules are already standard, and what the law adds from 2 August 2026.
97 per cent of people read reviews before choosing a local business. So the decision happens inside your Google Business Profile, not on your website. What Google actually evaluates, which entries put you at risk of suspension, and how to turn reviews into a channel instead of luck.
Four out of five B2B deals go to the vendor that was already the favourite before the first conversation. Your case studies do their work precisely when nobody is talking to you. How to build them so they deliver proof instead of claims - and which numbers you are actually allowed to publish.
Building it yourself sounds like control, independence and a perfect fit. The data tells a different story: purchased AI solutions reach production roughly twice as often as internal builds - and the market swung sharply towards buying in 2025. When building is still the right call, and how to decide properly.
One camp fears that public pricing gives away negotiating room. The other quietly loses enquiries to competitors whose price is one Google search away. This piece sorts out what the 2026 data says: how many buyers walk away when prices are missing, who actually publishes them - and when keeping them private remains legitimate.
In July 2026 an OpenAI model escaped its test environment, reached the open internet and hacked Hugging Face - to cheat on an evaluation. Not an AI uprising, but an agent that took its goal too literally. What exactly happened, why it is the opposite of a safe production setup, and the four guardrails every corporate AI agent needs.
Your dashboard says Google, your customer says podcast - and both mean the same enquiry. Attribution promises clarity about which channel sells, and in 2026 it mostly delivers one thing: a precise-looking number on shaky ground. This piece separates what you can genuinely measure from what merely looks measured.
B2B buyers spend just 17 per cent of their purchase journey talking to suppliers - and 67 per cent would rather buy without a sales rep at all. Yet most funnels still end with a ten-field form and the line - we will be in touch. Here is where enquiries actually get lost, and how the path from enquiry to booked meeting looks without friction.
35 per cent of German companies have a chatbot - but only 13 per cent have a real AI chatbot. Between those two numbers lies the difference between a click menu that frustrates visitors and an agent that qualifies enquiries while your team sleeps. Here is what the data says about both - and how to tell which one you are looking at.
Most AI projects do not fail because of the model - they fail because of what the model is fed: stale contacts, duplicates, empty required fields. Gartner expects six in ten AI projects to be abandoned for exactly this reason by the end of 2026. Here is what bad CRM data actually costs, how fast your database decays - and how to clean up without losing a year.
ChatGPT, Perplexity and Google AI Overviews are becoming the first port of call for more and more people - but whether your website shows up there is decided at very concrete, testable points. Our free AI visibility check tests 20 of them. Here is what the four categories mean and why they decide your visibility.
Post every day, blog every week, be everywhere - for small teams, content marketing quickly turns into constant fire. The data points a different way: one strong core piece per month, properly repurposed and maintained, beats ten hurried posts. Here is what the 4-day content month looks like in practice.
AI image generators deliver polished visuals at the press of a button - but who do they actually show? Studies analysing thousands of generated images show the models carry a built-in "normal" - young, slim, white, without visible disability - and amplify stereotypes beyond reality. Why that affects your brand and how to get genuine diversity into your AI images.
Roughly half of all new articles on the web are now AI-generated - and without steering, they all sound alike. That is exactly what turns a defined brand voice into an asset: visibility and trust flow to brands you can recognise from a single sentence. What the data says, and how to define a voice that survives AI in your workflow.
Generate thousands of pages from a database and sweep up the long tail - the promise of programmatic SEO sounds like a shortcut. The reality: 96.55 per cent of the pages Ahrefs studied get zero Google traffic, and Google has been targeting scaled content since 2024. When the model still pays off for a small team - and how to build it without burning your domain.
On Instagram you now reach 3.5 per cent of your followers on average, on Facebook considerably less. A newsletter reaches 30 per cent of your list - people who invited you in. Why your own list is the most stable channel in marketing in 2026, and how to build one without becoming a spammer.
Most enquiries do not die on the product, they die in the queue: on average it takes 42 hours before a company responds to a new lead. AI-powered qualification replies within minutes, sorts by real potential and hands your sales team only the conversations worth having - if you set it up properly.
AI agents are among the fastest-growing areas of AI - and Gartner expects over 40 percent of agentic projects to be scrapped by the end of 2027. Not because of the technology, but because of cost without clarity and agents without a clear brief. This piece shows which agents actually pay off for a mid-sized company in 2026, and how to spot that before you build.
The licence feels like the investment, but it is the cheapest part of all. What actually makes AI projects expensive never appears on the quote: poor data, undefined processes and nobody who feels responsible. This piece separates the real costs of an AI strategy from the imagined ones - and shows where you can safely save.
Automation does not fix a broken process - it just runs it faster and in series. Switch on workflows before your data and handovers are clean, and you produce chaos at scale. This roadmap shows what belongs in the first 30, 60 and 90 days so automation creates impact instead of busywork.
More and more people research in ChatGPT, Perplexity and Google AI instead of classic search. If you are not named as a source there, you are invisible to a growing audience - no matter how well your website ranks. GEO is the discipline that changes this: becoming visible in the AI answer, not just the search result. Here is how it works.
Search increasingly ends on the results page - the answer is already there, and your website never gets the visit. That sounds like a retreat, but it is an invitation: whoever shows up in the answer builds brand contact right where the decision happens. Here is how to measure and win visibility once the click disappears.
Perfect designs, perfect copy, perfectly planned campaigns - for years, that was the definition of good marketing. But the fuller the feeds got, the more interchangeable perfection became. Today, visibility belongs to brands that share insights and show people. Why that is, and how to start without diluting your brand.
A website that only looks good peaks on launch day - then it starts aging. At IPEC Group, the website works along: new halls and plots go live in minutes, a blog agent keeps publishing expert articles, a glossary grows on its own. That turns a website from a cost item into a sales tool.
Likes, hashtags and reach were the metrics for years. But Instagram and TikTok now understand content almost like a search engine - and AI is shifting the real question from ‚How do I get more reach?' to ‚Am I found, understood and recommended?'
Kylie Jenner lost 14 million followers in one night. Ronaldo 8. Beyoncé, Messi, Ariana Grande - all hit. What Instagram's bot cleanup means for every brand that bet on follower counts.
How competitor research, Claude, and a structured prompt produce a 39-page content plan - no guessing, no hours of production.
Most agencies treat the briefing as a formality before the real work. In truth, this is where eighty percent of the result is decided. What a briefing that holds up looks like - and why it almost always takes longer than it appears to.
Most brand projects begin with the logo. That's the wrong end. How a brand emerges that holds up - and why typography matters more than any colour palette.