Three Concept Case Studies

Devtool marketing sites are written for the person who signs. They’re read by the person who decides. Those are different people, and the site optimizes for the wrong one.

An engineer evaluating an API does one thing: tries it. Every section standing between the landing page and the first working request is a tax on that. Most devtool sites charge that tax three or four times — hero, social proof, feature grid, testimonial carousel — before revealing an install command.

The result is a site that reads well to a VP and gets skimmed by the only person whose opinion converts.

Invert the hierarchy. The quickstart is the homepage. Marketing copy earns its place only where it survives sitting directly next to working code.

Everything argued has to justify itself against something executable on the same screen.

Standard hero + logo wall + feature grid. Rejected because a logo wall is an unverifiable claim, and this audience discounts unverifiable claims by default. Worse: it’s the exact layout every competitor ships, so it differentiates nothing while costing above-the-fold space.

Interactive playground / embedded sandbox. Genuinely good, and rejected on cost. It’s weeks of build time to reduce time-to-first-success by less than showing a copy-pasteable curl command does in one afternoon. Right idea, wrong sequencing — this is a phase-two decision, not a launch decision.

Video demo above the fold. Rejected. It asks for ninety seconds of attention before delivering anything, from a reader who’ll give thirty.

Voice: Imperative mood, second person. No adjective that can’t be measured. “Fast” is banned; p99 under 40ms is not.

Type as a contract: Anything set in IBM Plex Mono is executable — copy it, paste it, it runs. Anything set in General Sans is argued, not proven. That rule has teeth. It means the type system tells the reader what they’re allowed to trust before they’ve read a word. Break the rule once and the whole system is a lie.

Structure: Quickstart → what it costs → what it doesn’t do → who it’s for. The “doesn’t do” section is deliberate. Stating limits early is the cheapest credibility available, and it filters out the wrong evaluator before they file a support ticket.

Primary: time from landing to first successful API call. Not signups. Signups are a vanity number in devtools; activation is the real one.

Secondary: quickstart copy-button click rate. It’s the cleanest available proxy for genuine intent.

Every practice site in this category is the same artifact: a photo gallery and a contact form. Price is hidden. Process is opaque. Timeline is unmentioned.

That produces a specific, expensive failure. The first call becomes a qualification call. The principal — whose time is the practice’s scarcest and most expensive asset — spends it discovering that a prospect’s budget was never viable. That conversation is being subsidized at the top of the funnel, repeatedly, forever.

The gallery makes it worse. Photographs of finished buildings compete on taste. Taste is unarguable. So the one thing that actually separates a good practice from a cheap one — how it thinks — is the one thing the site hides.

Publish the fee structure and the reasoning behind it. Move qualification upstream of the call, and make the thinking the exhibit.

Transparency here isn’t an ethics play. It’s a scheduling decision with a P&L attached.

Case-study-heavy portfolio site. Rejected because everyone has one, and finished-building photography systematically hides process. The photo is the least differentiated asset a practice owns.

Discovery-call funnel. Rejected on unit economics. It converts the principal’s hours into unpaid qualification labour at the widest part of the funnel. Any site that increases call volume without increasing call quality makes the practice poorer.

Full radical transparency — publishing every past project’s final cost. Tempting, rejected. Final costs are distorted by client-side variables the practice doesn’t control, so publishing them invites unfair comparison and probably breaches confidentiality. Publish the structure and the logic. Not other people’s numbers.

Navigation: “What this costs and why” as a top-level item. Not in the footer. Not inside an FAQ. The most avoided question in the category becomes the most visible link on the site.

Voice: Declarative and numeric. No euphemism. “We’ll tell you if your budget doesn’t work before you pay us anything” is the tone target — a sentence that’s useful and slightly uncomfortable, which is the combination competitors won’t copy.

Imagery inverted: Drawings over photographs. Plans, sections, sketches, redlines. Process artifacts, not glamour shots. If the differentiator is thinking, the site should show thinking.

Ratio of total inquiries to qualified inquiries. Total inquiries falling is an acceptable, expected outcome. The ratio improving is the goal.

Secondary: time on the fees page before contact. High time there followed by contact suggests the page is doing the qualification work it was built to do.

DTC product pages share one grammar: gallery, review stars, urgency banner, discount code. Because the grammar is identical across the category, the category is visually indistinguishable — and when nothing distinguishes, competition collapses to price.

That’s a trap for a small manufacturer. It cannot win a discount war against a company with better margins and worse products.

There’s a second, quieter problem. Pages that maximize impulse purchase also maximize returns, because they set expectations the product was never built to meet. Returns are where a small hardware brand’s margin actually dies.

The product page is an argument, not a gallery. State plainly what the product refuses to do.

Constraints are the only claim a competitor can’t copy without changing their product. Everything else — photography, story, tone — is copyable within a quarter.

Lifestyle-photography-led direction. Rejected as the most-copied and least-verifiable layer in the category. It’s also the most expensive to produce, which makes it the worst return on spend available.

Founder-story-led page. Rejected because the story is unfalsifiable and doesn’t survive a comparison tab. A customer with three tabs open is comparing specs, not narratives.

Review-volume-led social proof. Rejected as the primary axis — a small brand loses a volume contest by definition. Reviews stay, moved below the argument, and filtered by use case rather than star count.

Spec-sheet forward. Measurable claims lead. Interpretive claims follow.

Type as a contract, again: mono for anything that is a number or a sourced fact. Sans for anything interpretive. The reader can tell what’s verifiable at a glance, before parsing a sentence.

“What this isn’t” block, above the buy button. Not below. Not in an FAQ. Placing it before the purchase decision is the entire mechanism — it converts a post-purchase disappointment into a pre-purchase filter.

Return rate, primarily. If constraint-forward messaging works, returns fall — because expectations were set accurately before purchase, not corrected after it.

Secondary: add-to-cart rate, and bounce rate from visitors arriving with comparison intent.