Ten One-Pagers: One System, Ten Unique Faces
In April 2025 a Kyiv clinic owner asked for ten landings for ten services. Same brand, ten search intents, ten ad groups. My first quote scared him. Ten custom designs meant ten times the wait. I redrew the offer: one system, ten faces. One conversion skeleton for all, ten unique heroes and palettes from AI, one deploy pipeline. He paid once for the system, a small fee per face. We shipped ten in nine days. Leads cost fell from $14 to $7 in three weeks because we kept the faces that earned and paused the ones that did not.
Screenshot: Figma file cover shows ten frames, each palette hex listed, deploy log shows 10 pages published 2025-04-18 from single repo.
The skeleton that stays
All ten pages share one skeleton I test before I branch faces. The skeleton lives in code, not in Figma.
- Hero with one promise, one form, one proof photo.
- Three bullets that name the outcome, not the feature.
- Price or starting price, with one clear next step.
- Proof strip: three reviews or one case number.
- FAQ that answers price, timing, and guarantee.
- Sticky CTA that scrolls with the reader.
I build this skeleton once in Next.js and Supabase, with a content table for hero, bullets, price, and FAQ. Changing the skeleton changes all ten. Changing the face changes one. I read the finished skeleton against CRO Audit Template: 12 Checkpoints That Move Revenue.
The face that changes
For each variant I prompt Flux and ComfyUI for a hero art trained on the service, not stock. One service gets a calm clinic hallway, another gets a close up of hands, another gets a night street for logistics. I keep the palette locked to three brand hues plus one accent per variant. Copy stays human: I draft with a local LLM, then rewrite each headline aloud and cut any adjective I cannot point to on screen. Ten pages, ten headlines, one skeleton. Readers see variety. I see one system. The form and the sticky CTA decide most of the numbers, and I build those two blocks from Agentic UI Patterns That Actually Convert.
How I keep cost minimal
The system costs one build. Faces cost prompt time and one review loop each. My Trello shows 6 hours for the skeleton, 1.2 hours per face on average. That math turns ten custom quotes into one system quote plus ten micro-fees. Owners of small budgets buy it because they see the per-face price, not a big retainer. I ship the first face in 48 hours to prove the loop, then one face per day.
How I pick the winner
I run all ten with small ad budgets for ten days. Each page logs visits, form starts, and bookings. After ten days I pause the bottom five faces and move budget to the top five. One clinic kept only three faces after two cycles. Their lead cost held at $7 while others kept paying $14 for one generic page. The system pays because it learns. Learning needs a pipeline from analytics to money, and I lay that out in CRO as a Pipeline: From Analytics to Revenue.
What you can copy this week
Pick one service and write its skeleton as six blocks on paper. Build it once. Generate three faces with different hero prompts. Ship the three in three days. Keep the one that brings the form.