Build three versions of the same page: instant static artwork, delayed high-resolution motion, and progressive enhancement with useful text and controls first. Run each under the same recorded network and device profile, randomize presentation order, and ask testers to complete one identical information-finding task.
The hypothesis is that perceived waiting and abandonment intent will rise when meaningful content is blocked, even if total load duration matches a progressive version. Collect objective milestones, task completion, a short perceived-speed rating, and optional preference; keep visual richness separate from whether the answer arrived.
A reproducible report publishes source builds, throttling profile, device details, timing method, task script, order, and all missing responses. Until that evidence exists, “people prefer” remains off the poster and the sensible default is to render meaning before the overture.
The fastest flourish is the one that arrives after the answer.See our performance budget