The old case-studies page was a 3x3 grid. Nine tiles, each with a gradient placeholder and three lines of copy. It looked fine. It also closed nothing.
I sat with the analytics for an hour. Heatmap showed people scanning the grid, hovering on one tile, then bouncing without ever clicking through. The tile read as a brochure ad, not as evidence.
So I rebuilt it around the principle that case studies should READ like the inside of a war room, not like a Behance shot. Each case got its own /case-studies/[slug] detail page. Hero image (real, not gradient). KPI strip with before / after / delta. Problem statement in three paragraphs. Solution stack as chips. Implementation breakdown as a week-by-week ordered list. A real founder quote pulled into its own block. Inline service-link chips. One discovery-call CTA at the bottom, no exit-intent pop-up.
The index page changed too. Same 9 tiles, but every tile now carried a real generated hero image (no more gradient placeholders), and the click target became the whole card, not a tiny "read more" link.
Two weeks of data in: click-through from /case-studies to /discovery-call moved from 4% to 11%. Time-on-page on the detail pages averages 2m 40s. Three of those clicks turned into booked calls. One booked call closed for an $8,500 build — which paid for the redesign and the next three months of n8n hosting.
The lesson isn't "detail pages convert better than tiles" — that's obvious. The lesson is: grids reward browsing, stories reward deciding. If your service has a 30-min discovery-call ask attached, you don't want browsers. You want deciders.
