Open any agency's case study page. Scroll to any project. You'll find a metrics grid somewhere between scroll-row five and seven — four little boxes with big numbers, usually color-coded, usually forgettable.
The grid is a compromise. It says: we shipped lots of things, and we want credit for all of them. The problem is nobody remembers a grid. They remember one number.
The magazine test
Pick up any decent trade magazine. Wired, Fast Company, Monocle. Flip to a feature. The headline is never a list. It's one line. One assertion. One hook.
Good case studies are features, not dashboards. They earn attention the same way magazine covers do — by committing to one thing.
If a case can't be reduced to one honest sentence, it isn't ready to ship as a case study. Ship it as a work tile and move on.
How to find your number
The rule is narrow: pick the one metric a buyer in your target category would care about on a Monday morning. Not the vanity one. Not the technical one. The one that makes their Monday easier.
For a healthcare clinic, it's probably bookings. For a B2B manufacturer, it's RFQs. For ecommerce, it's revenue or conversion rate. For SEO work, it's rank movement or organic sessions. The industry tells you which lever matters most.
Once you have the lever, add the constraint: time. "47% more bookings" is weak on its own. "47% more bookings in six weeks" is a promise.
Where it goes
In the hero. Right below the project name. Same size you'd give a magazine subhead. Typography does the work — no badge, no widget, no grid.
On our own case studies, we typeset it at 5vw in Certia 500 with a coral-highlighted keyword and a coral period at the end. It reads like a line from an art-direction monograph:
47% more bookings in six weeks.
That's the whole case, compressed into one typographic moment. Everything else on the page is evidence for that sentence.
What this rules out
The rule has teeth. It forces you to skip the cases that don't have a real number yet. It also forces you to tell clients the truth when their results weren't what you hoped.
We've held projects out of the public case-study library for exactly that reason. The work was good. The numbers weren't there yet. Shipping them would mean fabricating a grid — and a fabricated grid is worse than no case at all.
One sentence. One number. One honest unit of time. Everything else is evidence.