B2B homepage positioning: do's and don'ts from 147 rewrites

Ten seconds. That is how long your B2B homepage positioning has to explain what you do, for whom, and why you. The Nielsen Norman Group analysed over two billion page visits: visitors who survive the first ten seconds stay. The rest are gone.

And every page gets checked. In a Wynter survey of 100 CMOs at software companies above 50 million in revenue, 100 percent said they look at the vendor's website before buying. Not most of them. All of them.

Most homepages don't survive those ten seconds. Not because of design. Because of messaging.

I wanted to know what exactly goes wrong. So I read the largest public dataset on this I know of: the before/after gallery of positioning agency Fletch PMM, 147 B2B homepages before and after their rewrite, with the agency's critique in bullet points. I opened all 147 pages, counted the critique on the 68 pages that carry it, and read 20 new homepages section by section. What follows is not the cases. It is the rules that follow from them, for your page.

What you'll learn

• The four faults that explain almost every one of the 147 rewrites, and how to fix them
• The nine-block skeleton the rewritten pages follow
• What belongs in the hero and why "AI" hurts there
• Every do and don't for homepage messaging in one table, plus three tests before go-live

My thesis: clarity is a decision, not a writing style

A B2B homepage doesn't fail on wording. It fails because nobody decided on one buyer, one problem and one enemy. Once that decision is made, the page almost writes itself, in the same order as 146 others.

It's an uncomfortable thesis, because it devalues copywriting and puts the founders on the hook. The proof comes in five steps: the fault count, the skeleton of the rewritten pages, the hero, the AI special case, and a second study that independently finds the same pattern. In between you get every rule as a table, and a test to run before the page goes live.

Why do B2B homepages fail on messaging, not design?

B2B homepages fail because they never chose one buyer and one use case. The four most common faults follow from that single omission: a message that is too broad, no differentiation, outcomes instead of explanation, no problem. Design can't rescue an undecided positioning. It only makes it prettier.

The numbers. Of the 147 cases, 68 carry the agency's critique as before/after bullets. I counted them. 38 pages say, in effect, "too broad, several audiences or use cases". 29 say "differentiation unclear, no enemy named". 28 say "outcomes without the how". 25 say "the problem is never stated". After that the curve drops steeply: jargon (9), feature wall (9), investor language (7), too long (7).

The top four are one fault seen from four sides. If you don't pick a buyer, you can't name an enemy, because every buyer has a different one. You can't describe a problem, because every segment has a different one. And you retreat into outcome language, because the how would differ per audience. "Ship better products faster" is the language of a page that didn't want to decide.

I've seen the cause across more than 100 B2B tech companies I've advised. "Too broad" is not a copy question. It's a founder decision nobody wanted to make, because every choice gives up part of the market. How that decision gets made is in my article on product positioning in B2B.

The four faults and what to do instead

Too broad. Root Signals addressed developers and enterprises at the same time, with the hero "Measure and Control GenAI". The rewrite cut the enterprises, focused on AI engineers and one concrete problem: evaluating LLM outputs. Ripe served growth, sales, RevOps and product. Afterwards: growth marketers only. For your page this means: pick the champion who feels the pain daily and can sign. Name them in the hero or the sub-line. The other personas get their own page or a section further down.

No enemy. RampMetrics got the note "doesn't name the enemy". The new page positions against homegrown tools and takes their weaknesses apart in detail. Hey Rosie compared itself to voicemail and answering services. The real alternative for its audience, small business owners, was: answering the phone yourself. The rewritten page says exactly that. April Dunford describes the status quo as the first competitor in any positioning: spreadsheets, manual processes, the tool you already have. By her numbers, enterprises lose 20 to 30 percent of deals to "no decision", not to competitors. For your page this means: name what the buyer does today instead, and take it apart in three concrete points.

Outcomes instead of mechanics. On UserEvidence the critique read: "leaves people wondering, but what does the product actually do?". It's the most quoted line across the 68 critiques. Passive described only the why, never the how. Instruqt never explained what the product does. For your page this means: capabilities first, because they make the outcomes believable. The outcomes move to the proof block, next to a customer name.

No problem. Polser had investor messaging on its homepage: very high-level, very broad, unclear what the product is and who it's for. Mo had a vision that won over investors and drove customers away. Pencil Spaces got a dedicated problem block so visitors know they're in the right place. For your page this means: give the problem its own block right after the logos, with three symptoms your champion would say in a stand-up. How a good pitch puts the problem before the solution is something I described in my article on the sales pitch after April Dunford. The same order applies on the homepage.

What does a B2B homepage that works look like?

A working B2B homepage follows nine blocks in a fixed order: a hero with buyer, use case and difference. Logos. The problem in the buyer's words. Why today's alternatives fail. How it works, in three steps with screenshots. Explicit differentiation, side by side. Proof with names and numbers. An FAQ that disqualifies. A closing CTA. Each block only works because the one before it has done its job.

I read 20 of the new homepages section by section, mostly from product management, developer tools and AI agents. The order is almost identical everywhere. A few examples to make the pattern tangible.

The hero names the difference, not the vision. Zeda: "Don't do product discovery in a planning tool. You need Zeda.io." Atolio: "The only AI-powered enterprise search engine that keeps your data in your cloud." The sub-line names the buyer or the use case. One primary CTA, one secondary.

The problem block comes right after the logos and speaks the champion's language. Root Signals: "Let's face it. You don't really know if your LLM features are delivering quality results." Then three numbered pains. Hey Rosie: "Currently, you have three bad options for tackling incoming calls." Zeda builds a without/with table with red crosses and green ticks.

Differentiation is laid out as a table. CommandBar: "How CommandBar is different from other Digital Adoption tools", six rows in the pattern "They do X, we do Y". Jit: "Everyone's slapping an AI label on old tech. Jit's Agents are different." The FAQ disqualifies on purpose: Scout asks "Who's NOT a good fit for Scout?".

What surprised me: the new pages are rarely shorter. Superhuman, SlimAI and CompanyCam got shorter. Apideck, Root Signals and Rosie got longer, because "how it works" and "why alternatives fail" were missing before. Length was never the problem. Order and decision were.

What belongs in the hero of a B2B homepage?

Three things belong in the hero: who the product is for, what you do with it, and the one structural difference to the alternative. Three lines, no metaphor, no outcome. If a buyer reads only the hero, they must be able to tell their team in one sentence what the product does.

Almost every before-hero in the 147 cases was a vision. Sprocket 365: "Rocket fuel for SharePoint". Scout: "Your toolkit for building scalable AI". Zeda: "Turn Voice of Customer Into Revenue Generating Products". Blue Triangle got the note "very clever, but not clear". Afterwards, Sprocket read "Finally, create a knowledge base in SharePoint" and Scout read "Automate sales workflows without bugging your dev team".

Six headline patterns keep recurring. "Don't do [job] in a [wrong tool]. You need [us]." "The only [category] that [difference]." "[Verb] [use case] without [pain]." "Finally, [the thing you couldn't do] in [the tool you already have]." "[Category] that get [work] done, not just [half-done]." "The first [mechanism] built for [narrow audience]." All six name a buyer or a use case and a difference. None names an outcome.

Anthony Pierri, co-founder of the agency behind the 147 rewrites, distinguishes two anchors here: category or use case. Figma positions in the category "collaborative interface design tool" and differentiates within it. Loom positions on the use case: "Skip the meeting, send a Loom." Both work. What doesn't work is a self-invented category no buyer has ever searched for. The hero is the shortest form of your one-liner: for whom, what, and the difference, in three lines, for exactly one person.

The AI special case: why "AI" in the hero hurts

Nine of the 147 cases are AI-native products from the last two years: Jit, Root Signals, Scout, Atolio, Passive, Anterior, DataFlint, Hey Rosie, Concierge AI. They make the same four faults as everyone else, plus a fifth: they sell "AI" as the value.

Pitch Kitchen calls this the AI parmesan effect. Pages with no AI mention in the hero score 24 out of 36 on average. Pages with three or more AI mentions score 12 out of 36. The more AI vocabulary, the less clarity. Among the 147 cases, Scout ("Your toolkit for building scalable AI") and Root Signals ("Measure and Control GenAI") show the same pattern. Passive explicitly wanted to distance itself from the "wave of AI tools" and "AI slop".

The correction follows one rule: say what the AI has read, what it does without being asked, and what it will not do. Jit after: agents that "act like product security engineers, powered by deep knowledge of your policies, compliance needs, and system architecture". And further down: "They aren't chatbots. They execute." Atolio after: data stays in your cloud, with a block on why foundation models, DIY storage and public-content tools don't fit the risk profile.

And the pages answer scepticism in the FAQ instead of ignoring it. CommandBar: "Is this 100% powered by AI?" Scout: "Will Scout replace our sales team or engineers?" That is buyer reality. Everyone has seen an AI label on old tech. If you want to use the terms cleanly, my list of AI GTM terms for 2026 has the definitions that are actually understood in the DACH market.

For companies in the DACH region there is one more layer. Data control here is not a compliance footnote but often the first objection. Atolio lifted its on-prem advantage from a subordinate clause into the H1. That was the entire rewrite.

Do's and don'ts for homepage messaging: every rule in one table

From the decision that comes first (audience) down to page mechanics (CTA). Every row is a pattern that was corrected several times across the 147 cases. Read the table top to bottom: each layer gets easier once the one above it is decided.

Layer	Don't	Do
Audience	Speak to two or three audiences on one page ("for product and engineering").	Pick one champion, name them in the hero or sub-line, write the whole page in their words. Other personas get their own page.
Use case	Describe a problem nobody owns ("improve alignment").	Lead with the use case someone wants to solve this quarter and recognises from their own to-do list.
Category	Invent a category no buyer has ever searched for, or float above all categories.	Anchor to a category with a budget line and name the difference inside it: "The only X that Y". Use the invented term only for the mechanism.
Enemy	Name no enemy, or pick a flattering one.	Name the real way it's done today (spreadsheet, homegrown tool, manual work, legacy suite) and take it apart in three concrete points.
Problem	Jump from the hero straight into features or benefits. "Teams struggle with alignment" doesn't count.	A dedicated problem block right after the logos, with three symptoms in the champion's language.
Hero	A vision, a metaphor or a hook that only makes sense on the second screen. An outcome in the H1 ("ship faster").	Three lines: what you do with it, for whom, the one structural difference. Capability or mechanism first, outcomes in the proof block.
Explaining the product	Explain only the why and leave the visitor asking "but what does it actually do?". Show every feature at once.	Three steps, each with a real screenshot: what you do, what the product does, what you get. Three to five capabilities, the rest on product pages.
Differentiation	Assume the difference is obvious once the product is explained.	A dedicated comparison block: "They do X / We do Y", with the mechanism that makes it possible. Take a stance.
AI	Sell "AI" or "agents" as the value. Data control as a compliance footnote.	Say what the AI has read, what it does unasked and what it won't do. Show the output, not the model. Answer "will it replace my team?" in the FAQ. Data control in the hero when the champion is technical.
Language	Vendor vocabulary or jargon that doesn't match the buyer. Hedging without a position.	The words the champion uses about their own job. Short sentences, concrete nouns (tickets, repo, sprint). One stance in the hero that some visitors will disagree with.
Structure	Paragraphs, the key point three scrolls down, a page so long nobody reaches the proof.	Headline plus one line plus one image per block. The headlines alone tell the story. The most differentiated feature first.
Proof	Broad claims without evidence, numbers that belong to nobody ("10x faster"). Leading with funding.	Every number next to a name, role and company, right beside the claim it supports. A few real quotes instead of a carousel.
Vision and deck	The pitch-deck story on the homepage. Investor vision instead of the customer's Tuesday. Mixing the paid product with open source or a side product.	The market thesis at most one sentence in the problem block. Commercial product first, the rest as "powered by". Say in the FAQ what you are not and who is not a fit.
CTA	Two equal CTAs split by persona. Hiding setup effort.	One primary CTA matching the buying motion (demo or trial), one secondary that lowers risk ("See how it works"). Time-to-value next to the button.

How do you test whether your homepage is clear?

Three tests are enough. Five-second test: after the first viewport, can someone say what you do, for whom and why you? Headline test: delete every sentence that isn't a heading. Do the headings alone tell the story? Enemy test: does the page say what the buyer does today instead, and why that's not enough? If one fails, the page isn't finished.

Then ten questions that follow from the 147 cases. A no on any of the first five means: don't go live.

1. Can a visitor name the champion from the hero alone, with role and team size?
2. Does the H1 or the line beneath it say what you do with the product, in the buyer's nouns?
3. Is today's enemy named and taken apart in at least three concrete points?
4. Is there a problem block with three symptoms the champion would say in a stand-up?
5. Is there a how-it-works sequence with a real screenshot per step?
6. Is the structural difference stated as a "they / we" comparison?
7. Does every number sit next to a customer name?
8. Does the FAQ answer who is not a fit?
9. Is there exactly one primary CTA per block?
10. Has every sentence containing "platform", "orchestration", "layer" or "AI-powered" passed a test against the buyer's vocabulary?

If you write your page not only for humans but also for agents, you have one more reason to be clear. In agentic GTM, buying agents read your homepage exactly like a CPO does: they look for buyer, problem, mechanism, difference. Vague pages fail both.

Does clarity on the homepage pay off measurably?

Yes, and a second study shows the same pattern as the 147 cases. In 2026, Pitch Kitchen scored 250 B2B homepages against 18 criteria: 190 of 205 companies name no concrete enemy. Every top-scoring page does. Scale-ups beat enterprise companies by 27 percent, because they can still afford an opinion.

Why this carries so much weight: according to Gartner, buyers spend only 17 percent of their buying time meeting with all potential suppliers combined. The remaining 83 percent, your website sells without you. Wynter shows what happens when it can't: buyers check five to eight vendors, three make the shortlist. Whoever doesn't find their problem on the homepage doesn't book a demo.

And the effect can be measured. For Cognism, according to Pierri, the switch from creative to clear led to 40 percent more conversion. One data point, not a study. But it points in the same direction as 147 rewrites and 250 scored pages: whoever decides gets understood. Whoever gets understood makes the shortlist.

Where the skeleton helps, where it doesn't, and the trap

✅ What shines. The order. Once buyer, problem and enemy are decided, you fill the nine blocks in an afternoon. The skeleton takes the taste debate out of copywriting and replaces it with a checklist.

❌ What doesn't shine. The skeleton without the decision. Nine blocks of "for teams of every size" are the same undecided page, only longer. And: the count comes from an agency's critique of its own cases. It shows patterns, not conversion data.

⚠️ Warning. The price of the decision is that part of your visitors will no longer feel addressed. That's not a side effect, that's the mechanism. Whoever reverses it in the review ("but we'll lose the CTOs") is back at case 1 of 147.

Back to the ten seconds. They are enough for exactly one story: one buyer, one problem, one enemy. The 147 rewrites tell that story 147 times. A company wanted to tell everyone everything. Then it decided. After that, the page almost wrote itself. A page nobody disagrees with hasn't convinced anyone either.

I'm not writing this from a distance. At Teklens we had thirteen product terms, two audiences and a vision hero. From the 147 cases we derived: one champion (product, not engineering), three elements in the hero, one named enemy (the chatbot that has never read the code), six blocks instead of twelve. Whether it works will show over the next few weeks. The vision hasn't died, it has moved: into posts and founder content. Why it does more there than in the hero is in founder-led content as a GTM advantage.

If you want to read analyses like this before they land here: subscribe to the newsletter. One operator note a week, data over opinion.