Analytics And Search Console Performance Audit

Analytics and Search Console Performance Audit

Human-facing summary

What this is: An audit that answers “Is organic search sending the right kind of traffic to the right pages?” using GSC page×query data (required for discovery) and optional GA4 landing-page data.

Two analysis modes (ask the user which they want):

| Mode | Best for | GA4 conversions | |------|----------|-----------------| | discovery_only | Services, publishers, low-volume sites — query intent and landing alignment | Optional (sessions for context only) | | discovery_plus_conversions | Sites with meaningful key-event volume — discovery plus visibility×value quadrants | Required when GA4 is provided |

Partial runs are valid: If only GA4 or only GSC is available, run the matching partial data mode; do not chase missing data.

Not this skill: Full hub discovery, keyword research, site remediation playbooks, or third-party rank data as a GSC substitute. Do not output numbered “what to fix on the site” recommendations — point to the Hub & Spoke Discovery playbook for next steps.

Packaging: Tier C ([write-a-skill](../../../utilities/write-a-skill/SKILL.md))

Reference: REFERENCE.md · Requirements: REQUIREMENTS.md · Examples: EXAMPLES.md

Force-load: "Use the analytics and Search Console performance audit skill to …"

Step 0 — Available data first (mandatory)

Before fetching anything, determine what the user already provided or explicitly asked you to pull:

| Available | Data mode | Action | |-----------|-----------|--------| | GSC page+query and GA4 landing pages | full | Score per analysis mode | | GA4 only | ga4_only | Ship partial audit — no GSC discovery | | GSC only | gsc_only | Ship partial audit — discovery from GSC only | | Neither | — | Ask once for at least one source |

Forbidden when a source is missing: npm installs (googleapis, etc.), DataForSEO/Ahrefs as GSC substitutes, inventing exports.

Completion criterion: handoff.mode matches available inputs; limitations[] states what is unknown.

Step 0.5 — Analysis mode (mandatory)

Ask once if not specified:

> Should this audit be discovery-only (query intent and landing-page alignment — best for services/low conversion volume) or discovery-plus-conversions (adds conversion/value quadrants)?

| User choice | inputs.analysis_mode | Requirements | |-------------|------------------------|--------------| | Discovery only / right traffic / no conversions focus | discovery_only | GSC page×query required for discovery summary; GA4 optional | | Both discovery and conversions / full matrix | discovery_plus_conversions | GSC + GA4; conversion events required when GA4 present |

Do not default silently — record the choice in handoff inputs.analysis_mode.

Completion criterion: handoff.inputs.analysis_mode is set.

Input policy

1. Domain — e.g. example.com (required)
2. Analysis mode — discovery_only or discovery_plus_conversions (required — ask in Step 0.5)
3. GSC — page + query export or equivalent (required for discovery summary)
4. GA4 — landing-page export (optional in discovery_only; required for full value matrix in discovery_plus_conversions)
5. Conversion events — 1–3 GA4 key event names (only when discovery_plus_conversions and GA4 provided)

Optional: date range, intended hubs (every URL must appear in report + discovery.intended_hubs[]), protected pages

Never infer domain, mode, exports, or conversion events from fixtures or prior runs.

Data modes vs analysis modes

Data mode (handoff.mode) — what sources exist: full | gsc_only | ga4_only

Analysis mode (inputs.analysis_mode) — how to interpret:

| analysis_mode | Quadrants (when GSC present) | Discovery block | |---------------|-------------------------------|-----------------| | discovery_only | Visibility only (high_visibility_unvalued / low_visibility_unvalued; indexing_noise → accidental_hub) | Required | | discovery_plus_conversions | Full matrix when GSC+GA4 | Required when GSC page×query available |

Quick start

node scripts/build-matrix.mjs \
  --domain example.com \
  --gsc path/to/gsc-queries.csv \
  --ga4 path/to/ga4-landing-pages.csv \
  --analysis-mode discovery_plus_conversions \
  --conversions demo_request,newsletter_signup \
  --intended https://example.com/,https://example.com/pricing \
  --date-range 2025-12-01/2026-05-31

Discovery-only (GSC required; omit --conversions):

node scripts/build-matrix.mjs \
  --domain example.com \
  --gsc path/to/gsc-queries.csv \
  --analysis-mode discovery_only \
  --intended https://example.com/pricing

Step 4 — Human report

Use handoff.discovery and handoff.pages[] as sources. Mandatory section order when GSC discovery is present (do not put strategic synthesis before evidence tables):

1. Banner — mode, date range, data sources
2. Query intent split — discovery.intent_split (clicks + impressions per intent)
3. Intended hub discovery table — every inputs.intended_hubs[] URL; verdict, top queries, GSC + GA4 session metrics
4. Misalignments — discovery.misalignments (state “none” if empty)
5. Indexing noise — discovery.indexing_noise_urls / indexing_noise flags (state “none” if empty)
6. Strategic read — one direct answer: Are keywords driving the right kind of traffic? Synthesize §2–5 only
7. Mode-specific sections (below)
8. Limitations
9. Closeout (one or two lines max) — playbook pointer and/or optional discovery_plus_conversions upgrade. No site fix list.

Quadrant counts are secondary in discovery_only.

discovery_only report (§7)

1. Visibility quadrant split — high_visibility_unvalued vs low_visibility_unvalued (+ accidental_hub only for indexing_noise URLs)
2. Top pages by GSC clicks — columns: URL, clicks, impressions, hub_type_hint, flags (from matching pages[] row)
3. High-impression / low-CTR landings (if any) — DIY or commercial queries with weak CTR; label by intent, not “accidental hub,” unless indexing_noise
4. GA4 landing context (if GA4 provided) — top pages by organic sessions when channel filter is available; otherwise all sessions; no converter leaderboard

Forbidden in discovery_only reports:

• Unicorn / hidden gem / prune / top-converters sections
• Numbered site recommendations (content plans, meta rewrites, internal linking tasks)
• Strategic read before §2 query intent split

discovery_plus_conversions report (§7)

1. Conversion context — site-wide key events; note when volume is low (quadrants are cohort-relative)
2. Performance quadrants — summary → accidental hubs → hidden gems → protected review
3. Intended hubs — quadrant + flags + absolute GSC/GA4 metrics

When key events are sparse: treat discovery sections (§2–6) as primary; quadrants are triage only.

Partial data modes

ga4_only: Sessions/intended hubs; no discovery block; upgrade line for GSC.

gsc_only: Sections §2–6 + visibility quadrants; upgrade line for GA4.

Step 5 — Handoff JSON

Emit handoff_version: "1.0" per REFERENCE.md. Include discovery when GSC page×query rows were parsed. Every inputs.intended_hubs[] URL must appear in discovery.intended_hubs[].

Done definition

• [ ] Step 0.5 — inputs.analysis_mode set from user choice
• [ ] Report section order §1–9 when GSC available
• [ ] Every intended hub in report table and discovery.intended_hubs[]
• [ ] Top GSC pages table includes hub_type_hint and flags
• [ ] No site remediation list in discovery_only
• [ ] node scripts/verify-handoff.mjs passes
• [ ] node scripts/verify-scorecard.mjs passes
• [ ] node scripts/verify-intent-classifier.mjs passes

Ship bar (Tier C)

node scripts/refresh-handoff-fixtures.mjs
node scripts/verify-handoff.mjs
node scripts/verify-scorecard.mjs
node scripts/verify-intent-classifier.mjs

Playbook consumer

[Hub & Spoke Discovery & Recovery](../../playbooks/hub-spoke-discovery-recovery/) — Phase 1a. Close reports with a pointer here — not inline remediation.

Maintainers

Follow [write-a-skill](../../../utilities/write-a-skill/SKILL.md). After scoring, discovery, or flag changes: refresh fixture, run all verifiers, update examples/SCORECARD-example-saas.md.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Roots-and-Fruit
• Source: Roots-and-Fruit/skills
• License: MIT

Install and usage instructions live in the source repository linked above.