Landing page work drove ₩900K–₩1.5M in monthly CPC savings on a ₩8M–₩15M/month Powerlink account. Ad copy iteration alone did nothing to the quality band.
I ran a 90-day split across two skincare supplement accounts in the same category, same AOV range (₩38,000–₩52,000). Approach A got dedicated landing pages per keyword cluster, sub-3-second LCP on mobile, keyword in H1, and structured product info (ingredients, dosage, KC certifications visible above fold). Approach B got only ad copy and bid adjustments — existing PDP, no structural changes. The hypothesis behind Approach B was that if quality score is largely locked in at the platform level, heavy landing page investment would show diminishing returns. That hypothesis did not survive contact with the data.
Approach A moved quality band from medium to high on 11 of 17 ad groups within three weeks. The fastest signal to register was review count visible above fold — session depth shifted inside a week. The slowest was URL structure matching the ad keyword — three-plus weeks, and the effect was inconsistent enough that I wouldn't prioritize it. What actually surprised me: Naver's algorithm appears to weight post-click engagement harder than I expected. Bounce rate dropped ~15% after adding structured product content, and that drop correlated with band improvements more cleanly than headline keyword match did. Approach B, by contrast, moved CTR 9–13% on brand-adjacent keywords through copy alone — but the quality band stayed flat. CTR without landing page alignment doesn't push you from medium to high. Naver appears to score those signals independently.
The more uncomfortable finding was about signals you simply cannot move. Historical account CTR is baked in at the account level. I inherited one account mid-flight after a previous operator's broad-match abuse had suppressed its history — identical landing pages, clean operation, and that account still dragged for the first 30 days. There's no workaround. Knowing this upfront would have saved at least one sprint cycle of misdirected optimization.
I wrote up the full breakdown — including the second structural signal that proved completely outside our control and how category competition density factors in — over on themedilog.
