Oceans Talent

Oceans Talent places pre-vetted remote executive assistants and operations specialists, called Divers, with US-focused companies. The site ran on Webflow with HubSpot for forms and meeting booking, Google Tag Manager and GA4 for analytics, Poptin for popups, and Notion as the editorial source of truth.

Performance. The homepage was failing Core Web Vitals, with a heavy payload and a slow first paint on mobile.

Untracked conversions. Three form systems ran side by side: HubSpot embeds, Webflow native forms routed to HubSpot, and Poptin popups, each with its own event model. The HubSpot booking embed redirected away before its submit callback could fire, so most booked meetings were never recorded in GA4.

Manual publishing and hygiene debt. Getting an article from Notion into Webflow's tabbed CMS structure was an hours-per-article manual rebuild: tab fields, TL;DR bullets, FAQ schema, author references, and internal-link checks, every post from scratch. A CMS template migration had left indexed legacy posts at 404 risk under a retired /blog-old path, older articles linked to a deprecated URL structure, the canonical domain was split between two domains, and category pages sat in draft where a link would have 404ed.

Performance Rebuild

Rewrote the Webflow foundation on a consistent Client-First class system. Cut render-blocking requests off the critical path, reformatted and resized image delivery, stripped unused JavaScript and CSS, and added explicit width and height to every image to stabilize layout.

Notion to Webflow Publishing Pipeline

A workflow using Claude with MCP connections to Notion and Webflow as the orchestration layer: write in Notion, push to Webflow as a draft, finalize in Designer, re-push as needed. Each article follows a fixed content contract (H2 sections mapping 1:1 to Webflow tabs, four TL;DR bullets, FAQ JSON-LD, relative internal links only) so automation stays deterministic. Along the way I solved a set of undocumented Webflow RichText quirks, from paragraph spacing that collapses without a zero-width joiner to an endpoint that auto-publishes drafts unless all field data ships in a single create call.

Unified GA4 and GTM Tracking Across Three Form Systems

For the HubSpot redirect problem, the fix was listening for the earlier, pre-redirect submit event and wrapping both callbacks in a dedupe object so either path fires exactly one GA4 event. Webflow native forms are targeted with a dedicated attribute selector, so unrelated forms on the same page produce no false positives. Poptin popups are handled in GTM via a click trigger. The tracking layer shipped with a GTM and GA4 handoff doc covering variables, triggers, tags, key events, custom dimensions, and a testing checklist, so the marketing team maintains it without me.

CMS Architecture and Content Cluster

Rebuilt the blog CMS on the same Client-First system with standardized rich-text styling, and launched a five-article Hiring and Onboarding cluster. The articles are cross-linked within the cluster, use relative internal links throughout, and share a common FAQ schema pattern.

Table Styling System

Consolidated cost-comparison and feature-matrix tables into three reusable CSS classes defined once in a global style block, so table embeds carry only semantic HTML, with a featured-row pattern that highlights the Oceans row in comparison tables.

SEO, Structured Data, and Site Hygiene

Every blog article emits BlogPosting schema, with FAQPage schema where relevant. The legacy /blog-old path now 301s to /blog, the canonical domain is unified on oceanstalent.com, and internal links pointing at the deprecated structure were swept. Behind this sits a repeatable link audit: pull all blog slugs from the CMS, pull all static page slugs from the Pages API, cross-reference every href, and flag anything pointing at a non-existent slug.

Case Study System Revamp

Redesigned the case-study collection end to end in about five working days: 12 new CMS fields designed and created via the API (SEO, metric, filter, curation, and date fields), a rebuilt detail template with a three-metric stat band that degrades cleanly when metrics are absent, and a full content backfill across all 16 published studies, with metric bands left honest (blank where the narrative had no real numbers). Every study now emits an Article, Review, and BreadcrumbList JSON-LD graph that passed Google's Rich Results Test with 4 valid item types and 0 errors (validated July 2026, re-run with matching results).

Metadata Audit and a Dedicated Meta-Title Field

A site-wide audit scored 52 pages on a 100-point, six-criterion rubric and mapped thin and duplicate metadata page by page, including a leaked "OG Description:" artifact on the homepage. Seven of the ten June proposals shipped and are live-verified. Structurally, the blog's <title> had been driven by the H1 field; a dedicated meta-title field now separates the two, so search titles and on-page headlines are tuned independently.

Gated Lead Magnet

Launched a gated 19-page PDF guide, the Chief of Staff Playbook, to production: a native form posting directly to the HubSpot Submissions API v3 (no embed), a return-visitor path that skips the form, and the PDF's embedded metadata scrubbed at the source before upload.

Content Growth

The blog grew from 42 to 60 published posts between late June and late August 2026, all through the pipeline; the August batch alone shipped five articles verified end to end on production (meta, H1, tab structure, embeds, JSON-LD, and 45 internal links).

Incident Recovery and Integrity Monitoring

In late August a routine audit surfaced that an AI-assisted brand-rename pass from early June had silently truncated 21 blog articles by 30 to 50 percent: the pass had regenerated whole rich-text fields instead of editing them, and everything it did not reproduce was deleted. I traced the damage signature, restored all 21 articles (18 from Wayback snapshots, 3 from a Webflow site backup), read every staged payload back byte-verified (60 fields), validated the recovered content by section structure and word count against its sources, live-verified the three backup-restored articles, published with zero errors, and disclosed the full incident and fix to the client the same day.

The permanent fix is a three-layer integrity system: a field-level CMS baseline across 92 items and 1,426 text fields that alarms on any shrunk or emptied field, a rendered-page baseline across 115 pages, and a monthly full-content archive in git as a restore source. A full link crawl the same week confirmed zero broken links across all 78 distinct internal targets.

• Homepage mobile Lighthouse Performance of 98, with LCP at 0.8s, TBT at 50ms, and CLS at 0.058.
• Accessibility 92, Best Practices 96, SEO 100 on the same run.
• Five-article Hiring and Onboarding cluster live, cross-linked, on a standardized tabbed template with BlogPosting and FAQPage structured data.
• Conversion tracking unified across all three form systems, including previously untracked HubSpot meeting bookings.
• Legacy 404 risk cleared: /blog-old redirects to /blog and the canonical domain is unified.
• Publishing runs through a documented Notion to Webflow pipeline the team operates without me.
• All 16 published case studies revamped onto a new template with validated Article, Review, and BreadcrumbList structured data: Rich Results Test, 4 valid item types, 0 errors.
• Blog grown from 42 to 60 published posts (late June to late August 2026) through the documented pipeline.
• Gated PDF lead magnet live, capturing leads straight into HubSpot via the Submissions API.
• Zero broken links across all 78 distinct internal targets (full crawl, August 2026).
• A 21-article content-loss incident found, fully restored, and closed in a single day, and a three-layer content-integrity monitor now guarding 1,426 CMS text fields weekly.