Marketing & comms · A4 portrait · v0.1.0
Case Study
A two-page editorial customer story: result-led headline, client profile, an oversized results band, the challenge in the client's own words, a numbered approach and a before/after table. Pick this for a B2B reference sales can hand over.
Paste the prompt into Claude (or any agent that can fetch URLs) with your content — you get back a finished, self-contained HTML file. Manifest ↗
Live preview — the example content ships with the template and is replaced by your content.
What an agent may change — and what content you should have ready. Everything outside these slots is fixed by design.
Slots
| Slot | Type | Req. | Example |
|---|---|---|---|
doc-title |
text | yes | Customer story — changeover time down 38% at Verhaeghe Profielen |
kicker |
text | yes | Customer story |
case-title |
text | yes | Changeover time cut by 38% — and Friday overtime disappeared |
client-name |
text | yes | Verhaeghe Profielen NV |
sector-line |
text | yes | Aluminium profile extrusion & anodising · family-owned since 1978 · supplies 400+ fabricators across the Benelux |
glance-industry |
text | yes | Aluminium extrusion and anodising |
glance-size |
text | yes | 240 employees, € 68 mln revenue (2025) |
glance-location |
text | yes | Kortrijk, Belgium (two press lines) |
glance-since |
date | yes | 8 Sep 2025 |
results-period |
text | yes | H1 2026 vs H1 2025 |
challenge-body |
html | yes | <p>We run about 1.400 profile dies across two presses…</p><p>The week was planned every Friday in a spreadsheet…</p> |
quote-text |
text | no | We had run SMED training twice before and both times it faded within a quarter. What changed is that the schedule stopped fighting us. |
quote-attribution |
text | no | Ilse Vandenberghe · Operations Director, Verhaeghe Profielen NV |
approach-body |
html | yes | <p>Nothing was installed in the first month. We timed changeovers with the die-setters…</p><p>Two of the three steps below are now run by the plant itself.</p> |
before-label |
text | yes | Before · H1 2025 |
after-label |
text | yes | After · H1 2026 |
results-body |
html | yes | <p>The table compares H1 2026 with H1 2025 on the same two presses, the same product-mix definition and the same measurement…</p> |
stack-body |
html | no | <p>Sequent Flow scheduling module (2 presses, 12 named users) · two-week on-site SMED baseline…</p> |
next-body |
html | no | <p>The same sequencing logic moves to the anodising line in Q4 2026.</p> |
cta-text |
text | yes | Short runs eating your press capacity? We will time your changeovers before we quote anything. |
contact-line |
text | yes | Joris Beekman · Sequent B.V., Eindhoven<br>joris.beekman@sequent.eu · +31 40 799 21 40 |
footer-line |
text | no | Verhaeghe Profielen NV · published with the client's permission, Aug 2026 |
Repeat: result-metric (2–4×)
One oversized headline number in the results band on page 1. Three is the designed count; four still fits, two looks sparse. Every metric must also appear as an outcome-row on page 2.
| Slot | Type | Req. | Example |
|---|---|---|---|
metric-value |
text | yes | −38% |
metric-label |
text | yes | Average changeover time |
metric-note |
text | yes | 74 → 46 minutes, timed die-out to first good profile across 1.248 changeovers. |
Repeat: approach-step (2–5×)
One step of the numbered approach on page 2, in the order it happened. Three is the designed count.
| Slot | Type | Req. | Example |
|---|---|---|---|
step-title |
text | yes | Two weeks on the floor, no software |
step-text |
text | yes | We stopwatched 46 changeovers on both presses with the setters and split every minute into internal and external work. |
Repeat: outcome-row (2–8×)
One before/after row in the results table on page 2. Include a row for every metric in the page-1 band, then the supporting ones.
| Slot | Type | Req. | Example |
|---|---|---|---|
outcome-metric |
text | yes | Average changeover time |
outcome-before |
text | yes | 74 min |
outcome-after |
text | yes | 46 min |
outcome-change |
text | yes | −38% |
Optional blocks
pull-quote— Oversized customer quote at the foot of page 1. Remove it whenever no real quote has been supplied and cleared — an invented quote is the fastest way to lose a reference customer.stack— Scope of work / tooling box on page 2. Remove when the engagement was a single product with nothing worth listing.next— 'What the client does next' box on page 2. Remove if the relationship has ended or nothing is planned; drop it first when page 2 runs long.
The authoritative fill instructions from manifest.json — the only place a template may instruct an agent (why).
Lead with the outcome: case-title states the RESULT, never the client's name. Every metric needs a basis and a period — metric-note carries the absolute before → after plus sample size, results-period carries the window ('H1 2026 vs H1 2025'); vague numbers destroy credibility. The table must reconcile exactly with the band: compute each outcome-change yourself as (after − before) / before ((46 − 74) / 74 = −38%; use 'pp' for shares: 61,2% → 68,0% = +6,8pp), and give every band metric its own row. Write challenge-body in the client's words, first person, about the situation before — not the vendor's pitch. Never invent a quote: with none supplied, delete the pull-quote block and ask the user to obtain one. Never name a real company without written permission — anonymise client-name to 'A Dutch logistics operator' when consent is unclear. European formats: 1.248, 61,2%, € 68 mln, '8 Sep 2025'. Step numbers are generated by CSS, so never type them. Keep it to EXACTLY two pages: delete every data-example block, then compress approach-body and step-text first, then drop next, then stack — never shrink the type. Missing details stay as [PLACEHOLDER] and are flagged to the user to fill via the built-in Edit button.
Published versions are immutable: /t/case-study@{version}/ never changes once live. /t/case-study/ always points at the latest.