Splitting one Spanish page into es-ES and es-MX: the numbers

Splitting one Spanish page into es-ES and es-MX: the numbers

Hypothesis: a single es page serving both Spain and Latin America was underperforming because of vocabulary and intent mismatch, not just hreflang granularity. Methodology: one publisher, one high-traffic guide, A/B-style by market over 10 weeks — we split es into es-ES and es-MX with localized copy (currency, idiom, examples), keeping a third Spanish-language market (es-AR) on the generic es as a quasi-control.

Findings, sub-pointed:
— Pure hreflang granularity (adding region codes without changing content) moved almost nothing in a two-week pilot: +2% clicks, within noise.
— Once we localized the actual copy under each tag, es-MX average position improved from 8.3 to 4.1; clicks from Mexico rose 61%.
— es-ES moved less: position 5.9 to 4.8, clicks +19% — Spain was already reasonably served.
— The control (es-AR on generic es) stayed flat, suggesting the gains were content-driven, not a sitewide tide.

Caveats: the Mexico query set had higher commercial intent, which inflates the click delta vs. impressions. We're measuring content localization as much as hreflang here — the two are entangled by design.

Conclusion: region codes are plumbing; the water is the localized content. The data suggests granularity without localization is near-useless.