The Research Hub went through a significant rebuild this release. Articles now render as a proper editorial experience — warm serif headlines, a sticky top-bar TOC that tracks your position as you scroll, and brand deep-dive cards that sit as a horizontal hero at the top of each pick rather than a cramped sidebar. The comparison table puts Tier first, platform logos appear in the column headers, and "Also Considered" entries now stack cleanly with the brand name and reason on separate lines instead of fighting a fixed-width layout.
Brand cards were showing 0/100 scores and blank platform data for most articles. The lookup now falls back through parent-brand and substring matching, so "Crayola Ultra-Clean Washable" correctly maps back to the Crayola score row. Every brand card and quick-pick now links through to its Visibility Scan Preview when one exists, showing real mention counts per platform instead of yes/no pills.
Two data quality fixes matter for anyone reading research articles. Brand domain resolution now appends topic context to the search query and filters out retailers, aggregators, and content publishers before picking a domain — the previous logic grabbed the top result blindly, which linked a vacuum-cleaner article's "Kirby" entry to a Nintendo fan site. Cross-topic contamination is also gone: a "best markers for kids" article was pulling in toy brands like Pokémon because they shared a "kids" tag with other synthetic scans. Brand rankings now come solely from that article's own research scans.
The weekly auto-refresh cron is off by default. With the pipeline now triggering lightweight brand scans per article, leaving the cron enabled across dozens of active articles would run up a meaningful tab silently. Set RESEARCH_AUTO_REFRESH_ENABLED=true to turn it back on. The sitemap build no longer fails on a transient database timeout — a connectivity blip during static export now produces an incomplete sitemap for that build rather than a failed deploy.