Paid Audit

Paid Ads Audit

Audit existing paid ad campaigns across Google Ads and LinkedIn Ads. Structured checklist covering account health, campaign structure, targeting, creative quality, tracking, and budget efficiency. Run quarterly or when performance degrades. No live API access required — works from campaign data provided by the user (screenshots, CSV exports, or verbal summary of metrics).

Live LinkedIn Ads data (optional): when the linkedin-ads MCP is authenticated, pull metrics directly instead of pasted CSVs — get_campaign_performance, get_creative_performance, get_audience_demographics, get_daily_trends, compare_performance. All reads are free and ungated; the MCP's write tools stay gated by .claude/rules/linkedin-ads-spend.md. Dormant until credentialed — see .claude/mcp/linkedin-ads/README.md.

Triggers

Run this skill when:

• Quarterly campaign health check is due
• CPL has spiked and the cause is unclear
• A new client is inheriting an account from a prior agency or in-house team
• Performance regressed and a structural review is needed before strategy refresh

Do NOT run when:

• Building a new campaign from scratch — use /paid-campaign-strategy
• Single ad copy is the issue — use /google-ads-copy or /linkedin-ads-copy
• Landing page is the issue — use /landing-page-audit

The Iron Law — voice-locked

EVERY FINDING MUST BE BACKED BY DATA OR FLAGGED AS UNAVAILABLE.

An audit built on assumptions is worse than no audit. If data is missing, say so. If a check can't be evaluated, mark it [UNAVAILABLE] and explain what data is needed.

No exceptions:

• "The account is probably fine" → Without data, you don't know. Flag the gap.
• "CPL seems high" → Compared to what? Cite the benchmark or mark as [ESTIMATED].
• "Just pause everything and restart" → Audit first. Most accounts need fixes, not demolition.
• "We can skip tracking" → Tracking is always the first audit category. No tracking = no optimization.

Fix-priority order — voice-locked

Always tracking first, creative last. Fixing creative without tracking is optimizing blind.

1. Tracking — foundation; nothing else matters without it
2. Structure — architecture; affects everything downstream
3. Targeting — audience; right message to wrong people = waste
4. Creative — messaging; right people, now optimize the message
5. Budget — efficiency; optimize spend after everything else is clean

This order is non-negotiable. A high-priority creative fix never outranks a tracking gap, no matter how impactful the creative looks. If tracking is broken, every "creative win" is unprovable.

Scoring bands — voice-locked

Each check is pass (1), warning (0.5), or fail (0). Category score = points earned / total possible. Overall score = average of category scores, weighted equally.

Inputs

Process

Five-phase flow: Platform identification → Google Ads audit (6 categories, 45 checks) → LinkedIn Ads audit (5 categories, 25 checks) → Cross-platform checks → Scoring + 30/60/90 day recommendations. Full flowchart, scoring methodology, and B2B SaaS benchmarks in the premium reference. The full check library lives in the premium reference.

Performance frameworks

The 75-check library finds configuration gaps; three diagnostic lenses read the numbers on top of it — the KPI ladder (Delivery → Engagement → Outcome, to localize where the leak is), spend concentration + wasted spend (the dollars sitting on ads with 0 conversions past the attribution window), and ad bucketing (drivers / promising / non-converters / too-early, with dollars attached). Every call obeys quantitative-evidence-floors.md — no "top performer" or "pause" verdict below the volume floor; state the count, name the lag. Full frameworks in the premium reference.

Anti-Hallucination Guardrails

1. Never invent benchmark figures. Cite the band from process.md or mark [ESTIMATED] with reasoning.
2. Never assume tracking is fine. If you can't verify GCLID flow / Insight Tag firing, mark Cat 1 / Cat 5 checks [UNAVAILABLE].
3. Never recommend pausing everything. Audit-driven fixes preserve the parts that work.
4. Never invent CPL / CAC / CTR numbers. If client didn't share, list as [UNAVAILABLE] and ask.
5. Never skip cross-platform checks. A platform-only audit misses budget-allocation and attribution problems.
6. Never crown a winner/loser below the volume floor. Per quantitative-evidence-floors.md — state the count behind any "top performer" or "pause" call. Below the floor (e.g. <3 conversions, <~100 clicks, before a full attribution window), it's a directional read with a "too early" caveat, not a verdict.

Quality

Pre-delivery checklist (categories scored, fixes ordered, recommendations actionable, benchmarks cited, gaps flagged), worked example, anti-examples (vague CPL, skipped tracking, generic creative recs, wrong fix order), and failure-mode triage in the premium reference.

Integration with Other Skills

Upstream (consumes)

Downstream (feeds into)