Performance is a design decision
You can't optimize your way out of a heavy design. Speed has to be chosen at the whiteboard, not retrofitted in the bundle analyzer.
Most performance work happens too late. The design is finished, the features are built, and now someone opens Lighthouse and starts negotiating with reality: compress the images, lazy-load the carousel, defer the scripts.
That work matters, but it's triage. The pages that score 99 aren't the ones with the cleverest optimizations — they're the ones designed light. The fastest carousel is the one the design never needed.
Choose your weight class early
Every design review should ask: what does this cost? A full-bleed video hero costs 4MB and your LCP. A third font weight costs a render-blocking request. A '3D touch' library costs 120KB of JavaScript for a hover effect CSS could approximate at zero.
This isn't asceticism. Rich visuals and fast pages coexist fine — gradients, SVG, transforms, and modern image formats are nearly free. The expensive things are usually the lazy things: raster images doing a gradient's job, JavaScript doing CSS's job, a library doing a function's job.
The budget is the brief
Set the performance budget before the first pixel: Lighthouse above 95, LCP under 2 seconds, zero layout shift. Then design inside it, the way an architect designs inside a plot of land. Constraints don't limit the craft — they are the craft.
Building something this essay touches on?
We're available for select engagements — let's talk.
Start a conversation