← Blog
Building Borker· 9 min read

Landing rebuild, part 3: The brief we threw out, and the rules that replaced it

By the Borker Team

Part 3 of eight on rebuilding our landing page. Part 2 is the decision that the page tells a story. The hub lists every part.

Starting it like a feature

I started the rebuild the way I would start a feature. A research pass first, with a researcher on each question.

The visitor's situationIn their own words, from a corpus of public reactions to a competing launch and from our own cold-email replies.
The categoryThe pages of every tool in it, section by section.
What convertsOn a product landing page today, with the sources.
What we may claimWhat the product actually does on production.
What we can measureWith the traffic we have.

Then three designers, each handed every sheet and told to produce a concept deliberately different from the others in how it shows the week. Then two critics scoring every concept on the same nine criteria, and an editor merging it all into one brief.

It came back as pages of text and tables. The research was good. The brief was too much to rule on.

The three concepts

Each concept came with three things.

DeliverableWhat it holds
Section ledgerFor every section: its purpose, the problem it answers, the mechanism, the evidence, the build cost
OutlineThe page at both widths
ArtboardsThree per concept

Concept A: The week

You have been burned by AI drafts and will not believe a headline. The first screen shows a real week from our own channels, drawn the way the app draws it, and the page proves every claim by pointing at a day in that week.

Concept A, The week: the hook and subhead on the left, and under it a seven-day grid of real posts from our own accounts, each card with its channel, its voice tag and where it came from

Concept B: The problem

You arrive doubting. The page opens on your situation and names your three objections out loud:

  • It will sound like AI.
  • It will rework what I already said.
  • I will not hand my account to something I cannot check.

It answers each with one of the three claims and a real post, and only then shows the week as the payoff.

Concept B, The problem: a situation line above the hook, then your three doubts said out loud, each linked to the claim that answers it

Concept C: In public

You have seen a hundred mockups of a perfect week. This page shows ours: every post that went out from our own accounts last week, none left out, each with its time and its voice, where it came from and whether a rule held it, and a link to the live post.

Concept C, In public: the hook on the left and one real post on the right, with four facts under it: who wrote it, where it started, when it went out, and whether a rule held it

The scores

The two critics scored on their own and picked the same one.

CriterionA, The weekB, The problemC, In public
Founder-centric354
Every element purposeful435
Claim-safe324
Honest proof435
Mechanism pushed down to the docs434
Mobile first534
Performance and accessibility445
Buildable in one package432
Distinct and memorable335
Total of 45, first critic342938
Total of 45, second critic353338

Why I threw it out

I read the brief on a laptop and could not rule on it. A visitor deciding whether to sign up would get even less from a page built on it. My verdict, in the words I used at the time: a decent starting point, emphasis on starting point.

The specific failures:

  • Too much text everywhere. Every concept was paragraphs. A visitor on a phone gets a wall.
  • Too static. Three pages of cards and tables. Nothing moved, nothing carried you.
  • Too founder-specific. All three pages were written for a startup founder. The people who have something to say and a social presence to keep alive are builders, artists, travellers, inventors, consultants. Founders are one of them.
  • Real data and names where there must be none. The concepts read our own workspace live and put real posts, real handles and my name on the page. A page that shows whatever went out last week shows next week's bad post too, before anyone has looked at it.

Concept C was the closest. Everything else was replaced.

What survived: the research

The research sheets are still the inputs for everything that came after. The mechanisms sheet in particular settled questions we would otherwise have argued about for a week, with a verdict and a reason per mechanism. A few rows:

MechanismVerdictReason in one sentence
Static headline, the hookuseShort and plain, and plain copy converts best in the largest benchmark we found
Product screenshot in the heroavoidShows the tool rather than the result, goes stale with the UI, unreadable on a phone
Real posts rendered as live textuseOur output is text, and real text is the fastest, most accessible and most honest proof we can show
Live generation demo in the heroavoidA cold generation cannot know the visitor's voice and would fail the hook in the first screen
A scripted demo presented as liveavoidBreaks the nothing-invented rule
Autoplaying video heroavoidHeavy, distracting, a pause-control duty, and text is a poor subject for video
AI voiceoveravoidA synthetic voice under the headline "Not like AI" argues against the page
Animation librariesavoidWeight for effects CSS already does
Logo strip, traction numbers, live signup popupsavoidWe have none we may show, the brand rules forbid the numbers, and the popups are pressure
Third-party post embedsavoidHeavy scripts, cookies and layout shift; render our own cards
Pricing on the landinguseBuyers ask for a public price, and the category hides it
Performance budget as an acceptance criterionuseSpeed lifted sales in every case study we could open
Every call to action opens account creationuseThe sign-in page leak outweighs any hero choice

The rules I wrote instead

Instead of a brief, a short list of rules, each one a reaction to a failure above.

  • Visual. The picture carries the meaning. A headline of six words or fewer and one line per beat. No paragraphs. A frame that needs a paragraph is wrong.
  • A story you scroll. One take, not ten cards. Scroll-driven transitions between beats, dynamic, alive. Every beat complete at rest, as a still. Reduced motion gets crossfades. The page works without the motion.
  • For any professional with a social presence. Not founders only.
  • No real data, no live read, no customer or founder name anywhere. Designed content only.
  • No graphics drawn or generated by agents in the final product. The reason is in part 5.
  • In stages, each a page I can scroll. Storyboard, then style, then the asset kit and the motion spec, then one act built for real, then the build. Nothing is built before the stage before felt right. A checkpoint is an artifact I scroll with a few sentences beside it, never a document.

The hard question stayed open: how do you show what Borker does without a demo of the product? The product's surfaces are not the point, and a screenshot of them is the product shot the research told us to avoid. The answer came from an unexpected place, and that is part 4.

The series: hub · part 1 · part 2 · part 3 · part 4 · part 5 · part 6 · part 7 · part 8

New here? Borker is the AI content engine for founders.

We learn your voice, read your sources, plan your week and write each post before it goes out, while you build the actual product.