Yetibeds
A large furniture catalogue that search engines and AI shopping assistants could not read.
Beds and furniture
This story describes what was shipped, not a measured lift. No conversion, traffic or ranking figures appear here — see the note on numbers on the case studies hub.
The problem
Platform beds, daybeds, upholstered storage beds — a deep catalogue, and structurally invisible to the systems that decide whether it gets shown.
Product pages were missing Schema.org Product structured data, so Google Shopping and AI shopping assistants had no machine-readable description of what was being sold. They were also missing H1s and barcodes / GTINs — the identifier a product feed is matched on. Without it, a bed is not a product any aggregator can place.
The performance picture was as bad in a different direction: no preload hint for the hero image, so the largest element on the page waited its turn in the network queue, and images without dimensions, which let the layout jump around as they arrived. Both are Core Web Vitals blockers. On top of that, slow product pages, CTAs with poor visibility — including the search button itself — and frustration clusters concentrated on product pages.
What we fixed
Applied across the catalogue and verified live:
- Around 60 page titles and 50 meta descriptions
- About 10 image alt texts
- Barcodes, closing the Google Shopping identifier gap
- App-embed schema
Structural fixes:
- Product Schema.org JSON-LD
- A missing product page H1
- An LCP preload hint for the hero image
- GEO fixes for AI-search readiness
- Theme audits
Friction fixes: search-button visibility, and a form fix.
Retention layer built: exit-intent rescue, cart rescue using a free-shipping progress bar plus cart reminders, email-capture popups, spin-to-win, an added-to-cart celebration, and product-page discounts.
Catalogue cleanup: 14 product updates executed through the agent, with more queued for approval, plus discounts.
How we fixed it
SEO Autopilot applied the fixes, one verified click at a time
Each write goes through Shopify's Admin API and is then read back to confirm it actually landed — and every applied fix stores its previous value, so any one of them can be undone individually.
The GEO audit flagged the AI-search gaps
The structured-data and readability gaps that decide whether an AI assistant can describe your product at all, as distinct from ordinary search ranking.
Seven CRO Reports ranked what was left
Funnel, frustration and performance issues, ordered — so the work queue was a sequence rather than a pile.
SmartNudge and DynoAgent did the building
The intervention suite came out of SmartNudge; the bulk product editing went through DynoAgent, executing on approval rather than autonomously.
The end result
The catalogue went from largely unreadable to structured and machine-readable for Google Shopping and AI assistants. The Core Web Vitals blockers — LCP preload, layout shift — were addressed. A complete retention layer was deployed. And the tedious part, product-data cleanup across a large catalogue, was done by the agent rather than by hand.
How DynoWeb helped
The SEO and performance team a furniture store cannot afford to hire.
The gaps here were not matters of taste. Missing structured data, missing GTINs and an unpreloaded hero image are objectively wrong, tedious to fix, and individually worth very little — which is exactly why they survive for years on a catalogue this size. Closing them is a week of specialist work, or a sequence of one-click verified fixes.
Features used
SEO Autopilot
Applied titles, descriptions, alt text, barcodes and schema — each read back to confirm it landed, each reversible.
Storefront Speed
Named the LCP and layout-shift blockers on the slowest revenue-exposed pages.
CRO Report
Seven audits ranking the funnel, frustration and performance work.
DynoAgent
Executed the bulk product updates on approval, not autonomously.