Marketing AdsPPCPaid Media

Paid Ads

Paid advertising campaigns on Google, Meta, LinkedIn, and more.

$skillfish add ~skills/paid-ads

Real Examples Get Started Models & Context Good to Know

Real Conversations

See how the Paid Ads skill responds to actual requests. These are representative examples of what your AI agent will produce.

You

We're spending $8,000/month on Google Ads with a 4.2% CTR but only 1.1% conversion to trial. CPC is $12. We're targeting keywords like 'project management software' and 'team collaboration tool'. Something is off.

Agent

Paid Ads Audit: Google Ads Account

The diagnosis: Your CTR is strong (4.2% is above SaaS average of ~2.8%), which means ad copy is resonating. The 1.1% trial conversion is the problem — this is a landing page and keyword intent issue, not an ad issue.

Root cause 1: Keyword intent mismatch "Project management software" and "team collaboration tool" are top-of-funnel research queries. These users are 60–90 days from buying. You're paying $12 CPC for browsers, not buyers.

Better keyword structure:

Campaign	Keywords	Intent	Est. CPC
Brand	[your brand name]	Purchase	$2–4
Competitor	[competitor] + alternative/vs/pricing	High	$8–15
Bottom-funnel	"project management software pricing" / "best PM tool for [niche]"	High	$10–18
Top-funnel	"project management software"	Low	$8–12 — reduce budget here

Root cause 2: Landing page mismatch If you're sending all traffic to your homepage, that's costing you 50–70% of potential conversions. Each campaign/intent tier needs a dedicated landing page.

Immediate actions:

Add negative keywords: "free", "open source", "template", "examples", "tutorial"
Shift 30% of budget from top-funnel to competitor + bottom-funnel campaigns
Build a dedicated trial landing page with one CTA

Expected outcome: Trial CVR should move from 1.1% to 2.5–3.5% within 45 days.

You

We want to run LinkedIn ads targeting VP of Engineering and CTO at SaaS companies with 50–500 employees. Budget is $5,000/month. We've never run LinkedIn ads before. Walk me through the setup.

Agent

LinkedIn Campaign Setup: VP Eng / CTO Targeting

Account structure:

Campaign Group: Engineering Leadership — Q2
  Campaign 1: Awareness (Sponsored Content)
  Campaign 2: Consideration (Document Ads)
  Campaign 3: Retarget (Sponsored Content)

Targeting settings for Campaign 1:

Job Title: "VP Engineering", "VP of Engineering", "CTO", "Chief Technology Officer", "Head of Engineering"
Company size: 51–500 employees
Industry: Computer Software, Internet, IT & Services
Do NOT layer on seniority — it restricts reach and LinkedIn's title targeting already captures seniority
Estimated audience: 80,000–150,000 (healthy for B2B)

Budget allocation ($5K/month):

Campaign 1 Awareness: $2,500 (CPM bidding, $45–65 CPM expected)
Campaign 2 Document Ad: $1,500 (CPC bidding, $8–14 CPC expected)
Campaign 3 Retargeting: $1,000 (website visitors + video viewers)

Creative specs:

Single image: 1200x628px, headline ≤150 chars, max 3 lines visible before "see more"
Lead gen forms: pre-fill removes friction — use for gated content offers

Realistic benchmarks:

CTR: 0.4–0.8% (0.4% is acceptable on LinkedIn)
CPL with lead gen form: $80–180
At $5K/month: expect 28–62 leads/month

First 2 weeks: Run 2 ad variants per campaign, kill the lower CTR one at 1,000 impressions each.

Get Started

How to install and use this skill in your preferred environment.

Skills are designed for AI coding agents (Claude Code, Cursor, Windsurf) and IDE-based workflows where the agent can read files, run scripts, and act on your codebase. Web-based AI can use the knowledge and frameworks, but won't have tool access.

Models & Context

Which AI models and context windows work best with this skill.

Recommended Models

Best

Claude Opus 4 Claude Sonnet 4 GPT-4.1 Gemini 2.5 Pro Grok 3 Kimi K2

Good

Claude Haiku 4.5 GPT-4.1 mini Gemini 2.5 Flash Grok 3 mini

Larger models produce more detailed, production-ready outputs.

Context Window

This skill's SKILL.md is typically 3–10 KB — fits in any modern context window.

8K Skill only

32K+ Skill + conversation

100K+ Skill + references + codebase

All current frontier models (Claude, GPT, Gemini) support 100K+ context. Use the full window for complex multi-service work.

Pro tips for best results

1

Be specific

Include numbers — users, budget, RPS — so the skill can size the architecture.

2

Share constraints

Compliance needs, team size, and existing stack all improve the output.

3

Iterate

Start with a high-level design, then ask follow-ups for IaC, cost analysis, or security review.

4

Combine skills

Pair with companion skills below for end-to-end coverage.

Good to Know

Advanced guide and reference material for Paid Ads. Background, edge cases, and patterns worth understanding.

Contents

Quality Score and Ad Relevance by Platform

Google Quality Score is a 1–10 diagnostic score per keyword, comprising three components:

Component	Weight (approximate)	What it measures
Expected CTR	~35%	Historical likelihood of clicks relative to ad position; benchmarked against similar ads
Ad Relevance	~30%	How closely ad copy matches the keyword's intent; checks headline and description alignment
Landing Page Experience	~35%	Page relevance to keyword, load speed, mobile usability, absence of excessive interstitials

Quality Score affects your Ad Rank (position) and effective CPC — a QS of 8 can outrank a competitor bidding higher with a QS of 4.

Meta Relevance diagnostics (no single score — three separate ratings in Ads Manager):

Quality ranking: creative quality vs competing ads for the same audience
Engagement rate ranking: expected engagement vs competing ads
Conversion rate ranking: expected conversion rate vs competing ads with the same optimization goal

LinkedIn relevance: LinkedIn does not publish a score equivalent. Relevance is reflected in CPM/CPC volatility — highly relevant ads for an audience see stable or declining CPCs over time; low-relevance ads see rising CPCs as LinkedIn optimizes delivery toward better-performing creative.

Bid Strategy Decision Tree

Automated bid strategies require conversion data to function correctly. Using smart bidding without sufficient conversion volume produces erratic results.

Strategy	When to use	Conversion volume threshold
Manual CPC	New campaigns, testing phase, very low volume	Any volume; you control bids directly
Maximize Clicks	Drive traffic when no conversion tracking exists	No conversions required
Maximize Conversions	Scale conversions within a fixed budget	15–30 conversions/month per campaign minimum
Target CPA (tCPA)	Optimize to a specific cost per conversion	30–50 conversions/month for stable performance
Target ROAS (tROAS)	E-commerce or high-value lead gen with revenue tracking	50+ conversions/month; revenue values must be passed

Common mistake: Switching to tCPA or tROAS before hitting the volume threshold. The algorithm enters a learning period with insufficient data, CPAs spike, and campaigns are paused prematurely. Use Maximize Conversions first to build the conversion history, then transition to tCPA once thresholds are met.

Audience Match Rate Issues

Customer list uploads consistently underperform expected match rates. Typical platform match rates:

Platform	Expected match rate	Common cause of low rates
Google Customer Match	40–60%	Unhashed emails, personal vs work email mismatch
Meta Custom Audiences	50–70%	Old email list, incorrect hashing format
LinkedIn Matched Audiences	15–40%	Work email required; personal emails don't match

Hashing requirements: All platforms require SHA-256 hashed email addresses. Hash must be lowercase before hashing — John@Example.com must be normalized to john@example.com then hashed. A common error is hashing before lowercasing, which produces non-matching hashes.

Improving match rates:

Upload multiple identifiers simultaneously (email + phone + name + ZIP) — platforms use probabilistic matching across all signals
Use work email addresses for B2B platforms (LinkedIn match rates improve significantly with work emails)
Refresh lists quarterly — email churn rates of 20–30%/year mean old lists match poorly
On Meta, enable Advanced Matching to capture hashed parameters client-side in addition to list uploads

Creative Fatigue Signals

Frequency thresholds vary by platform and campaign objective, but the underlying signal is the same: CTR declines as the same user sees the same creative repeatedly.

Platform	Frequency warning threshold	Observation window
Meta	2.5–3.5x for cold audiences	7 days
LinkedIn	4–6x before significant CTR decline	Campaign lifetime
Google Display	3–5x per week	7-day rolling

Reading a frequency-to-CTR decay curve: Pull a report with frequency as a dimension and CTR as a metric, segmented by week. A healthy creative shows flat or slowly declining CTR as frequency increases from 1x to 3x. Fatigue looks like a steep CTR drop between frequency buckets (e.g., CTR halves from 2x to 4x frequency). When you see this pattern, creative rotation is overdue.

Response to fatigue: Refresh creative assets before frequency-forced performance decline. On Meta, maintaining 3–5 active ad variations per ad set with automatic creative optimization reduces fatigue systematically. On LinkedIn, plan a creative refresh every 4–6 weeks for evergreen campaigns.

Attribution Window Trade-offs

Attribution windows determine which ad interactions receive credit for a conversion. Different windows produce materially different ROAS numbers from the same underlying data.

Window	What it counts	Best for
1-day click	Conversions within 24h of a click	Short purchase cycles; direct response
7-day click	Conversions within 7 days of a click	SaaS trials, considered purchases
28-day click	Conversions within 28 days	High-consideration B2B, long sales cycles
1-day view-through	Conversions within 24h of an impression (no click)	Brand awareness measurement; easily inflated

Why windows produce different ROAS numbers: A user who clicks an ad on Monday and converts on Sunday is counted under 7-day click but not 1-day click. View-through conversions are the most inflationary — users who saw an ad and later converted organically get attributed to the ad.

Choosing a window: Match the window to your actual sales cycle. If your trial-to-paid conversion happens within 24 hours, a 1-day click window is accurate. If users evaluate for two weeks before converting, a 7-day window undercounts. For B2B with 30-60 day sales cycles, 28-day click is more representative but still misses offline conversions — supplement with CRM-based attribution for deals over a threshold value.

Cross-platform double-counting: Each platform applies its own attribution window and takes credit independently. Meta, Google, and LinkedIn will each claim full credit for the same conversion. Use a single source of truth (GA4, your CRM, or a third-party MTA tool) for budget decisions rather than summing platform-reported conversions.

Ready to try Paid Ads?

Install the skill and start getting expert-level guidance in your workflow — any agent, any IDE.