Marketing needed twelve campaign pages a month. Every one queued behind a designer. So I stopped designing pages and designed the thing that makes them.
Kushals runs campaigns constantly — festival drops, category pushes, influencer collabs, clearance. Each one wanted its own landing page, and each one arrived as a Slack message with a deadline attached.
The real cost wasn't the design time. It was the queue: a page took roughly ten days from brief to live, most of it spent waiting. Marketing planned around that lag, which meant campaigns got scoped down before anyone drew anything. And because every page was bespoke, the brand drifted a little further each month.
I pulled every campaign page from the previous year and took them apart. Under the custom art direction, the same structures kept recurring: a hero, a category grid, an offer strip, a product rail, an editorial band, a trust row, a footer CTA.
That became the brief. Not "build a page builder" — build these seven blocks, with the variations that actually got used and nothing else. A generic builder would have recreated the original problem in a new tool.
Marketers can change anything that's reversible — copy, imagery, product selection, order of blocks. They cannot change anything that compounds: type scale, spacing, colour, button treatment, grid. Those are locked at the block level, so a page assembled by someone who has never opened Figma still looks like Kushals.
The first version worked and almost nobody used it. Marketers kept sending briefs, because a brief was familiar and the builder was a new thing to be bad at in front of colleagues.
What fixed it wasn't UI. It was seeding the tool with six pre-assembled campaign templates so the first action was edit, not create; sitting with two marketers while they shipped a real page under deadline; and getting the team lead to publish their own page first. Adoption went from a handful of pages to 80% of campaign traffic within two months.
I'd build the templates first next time. We shipped a tool and then discovered that the tool wasn't the product — the starting points were.
The lock-down decision still holds up. Every request to "just make the heading size editable" would have cost more than the ten days it saved, and saying no to those is most of what kept it working.