CavCore console · CavBot analytics

Guardian view of your routes & SEO

CavCore Console is the dashboard behind CavBot — an event-first map of pageviews, 404 recoveries, SEO structure, referrers, and runtime feel. Every card on this screen is backed by append-only events and derived tables ready for a future AI layer.

Backing tables: events, daily_page_aggregates, seo_snapshots, performance_samples, insights.
Project cavendishpierrelouis.com
Window Last 30 days
Focus All components




Filters

Choose which project, window, and signal CavBot should show here. These knobs drive every card on the console — from 404 density and SEO to referrers and deploy markers.


Quick focus




Overview

Sessions, events, and 404 density

A clean roll-up of CavBot’s event stream for this project — every number on this row comes directly from events, sessions, and daily_page_aggregates.


Total events
284,930
All components · last 30d.
Sessions
19,286
Distinct session_key per project.
404 sessions
1,042
Sessions containing 404-control-room events.
Badge interactions
3,117
cavbot_badge_click / hover events.
30-day route & 404 trend
Sessions 404 views
D1
D2
D3
D4
D5
D6
D7

CavBot never stores PII — trends are built from anonymousId, sessionKey, route, and component, not from user accounts.





SEO & indexability

Route-level SEO snapshots

CavBot takes periodic seo_snapshots per page — title, description, canonical, indexable flags, headings, word count, and social tags.


Total pages · 143 Indexed pages · 127 Pages with critical SEO issues · 9 Pages with minor issues · 24

  • Duplicate titles detected on 3 blog posts.
  • Missing or thin meta descriptions on 6 marketing pages.
  • Canonical mismatch on /404 variant — points to outdated route.
  • 2 high-traffic pages marked noindex unintentionally.

These rows feed insight types like seo_missing_meta, duplicate_title, and non_indexable_critical_page.

Runtime feel

Performance snapshots per route

From performance_samples, CavBot tracks LCP, TTFB, CLS, and FID per page and device, so slow-but-important pages float to the top.


Average LCP (desktop)
1,980 ms
Across all monitored paths.
Average TTFB
210 ms
Origins within target budget.
Pages above LCP budget
5
Flagged for insight slow_critical_page.
Pages with unstable layout
2
CLS > 0.1 on mobile.
Structure & internal links

Canonical map & weak routes

CavBot uses the pages table, internal referrers, and sitemap data to detect orphan and under-linked pages.





Orphan pages
3
Appear in sitemap, but never linked internally.
Weakly linked routes
7
Important pages with < 3 internal links.
Sitemap mismatches
4
404 when visited from sitemap.xml.
High-value funnels
2
Tagged with importance_score > 0.8.

This is where CavBot earns its “guardian angel” role: flagging pages that should carry more structural weight.





Referrers & campaigns

Who sends traffic into broken routes

CavBot’s referrer_aggregates table rolls up traffic by domain and UTM campaign — highlighting sources that over-index on 404s or thin pages.


Referrer
Sessions
404 hits
404 rate
google.com
8,921
143
1.6%
linkedin.com
1,284
57
4.4%
newsletter.cavbot.dev
622
2
0.3%
unknown / direct
5,131
190
3.7%

Campaign health is computed the same way — aggregating UTM source / medium / campaign with error rates and top landing pages for each.


Campaign 404 performance
Last 30 days · utm_campaign
launch_rinasce_collection
sessions 2,184 · 404 rate 1.2%
newsletter_november_rollout
sessions 1,046 · 404 rate 0.9%
linkedin_brand_awareness
sessions 712 · 404 rate 3.8%

Bars show relative traffic volume per campaign. Color intensity and badges reuse the same referrer_aggregates and insights schema as the table above, so CavBot can talk about which campaigns are healthiest — or sending visitors into dead ends.

Guardian insights

Issues CavBot is watching

The insights table stores typed, severity-ranked findings that future AI layers can “read” and talk about — each with structured context.


404_spike
High
404 spike after last deploy on /blog
312 cavbot_404_view events in 24h · mostly from internal links on /blog. Suspected: missing redirect for an older article slug.
page_type: 404-control-room component: core-tracking status: open
seo_missing_meta
Medium
Key marketing page without meta description
/pricing appears in top 10 landing pages but has no meta description. Derived from latest seo_snapshots.
seo high-traffic status: open
slow_critical_page
Medium
Onboarding route is slower than budget
Average LCP 2,420ms on /onboard with high importance_score. Based on performance_samples aggregates.
performance importance > 0.8 status: open

Alerts for these insights (webhook, email, dashboard banners) are stored in the alerts table, keeping a clean history of when teams were notified.





Deploy markers

Align incidents with releases

CavBot’s deploy_markers table helps you line up 404 spikes, SEO regressions, and perf changes against your own deploy timeline.


Deploy #127 — Blog routing refactor
2025-11-28 · 404 spike detected on /blog within 2h of deploy.
Deploy #126 — 404 control room update
2025-11-24 · Catch rate improved by 11.3% after UX change.
Deploy #125 — Image optimizations
2025-11-19 · LCP on /home dropped from 2,380ms to 1,740ms.
Data & privacy health

Event-first, privacy-respectful

CavBot’s analytics layer is designed to be safe for real production websites: no emails, no names, no long-term IP storage — only coarse, anonymous signal.


  • Anonymous visitors only: CavBot uses project-scoped anonymousId, never raw user IDs.
  • No PII in payloads: Events avoid emails, names, or custom identifiers by default.
  • IP handling: Optional, truncated, and used only for aggregated geography if enabled.
  • Append-only events: Raw events can be archived after N days, while daily aggregates and insights are retained longer.

PII risk
Low — payload schema excludes names, emails, and raw IPs by design.
Aggregation coverage
Most metrics come from daily_page_aggregates and referrer_aggregates, not raw events.
Retention horizon
Raw event retention tuned to 90 days with longer-lived, anonymised rollups.

This project view is intentionally structured so a future CavBot assistant can sit on top of events, seo_snapshots, performance_samples, and insights — without ever needing to see personal identities.