Install any skill in seconds. Free to start, no credit card required.
Get Started Free →When the user wants to set up, optimize, or manage Google Ads campaigns. Also use when the user mentions "Google Ads," "Google Search Ads," "PPC," "SEM," "PMF testing with ads," "test product-market fit," "Responsive Search Ads," "RSA," "Performance Max," "Quality Score," "keyword bidding," or "Google Display/YouTube ads." For paid mix, use paid-ads-strategy.
.claude/skills/mkurman-google-ads/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 116% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -2% | 0% |
|------|--------|--------------|---------| | PMF testing | Pre-PMF; validate idea before building | $47–500; start small | Simple LP: headline, benefits, problem solved, CTA ("Join Waitlist," "Get Early Access") | CTR, sign-up rate, bounce rate; low CTR/high bounce = messaging/positioning issue | | Conversion-driven | PMF validated; commercialization | Scale; ROAS target | Full funnel; ad-to-page alignment | ROAS, CAC, conversion rate |
PMF testing: No full product needed. Build landing page with Unbounce, Carrd, or Webflow. Run ads to relevant search terms; measure clicks, engagement, signups. Test messaging (e.g., "Fastest App for Freelancers" vs "Simplest Time Tracker for Teams"), pricing (different price points in ads/LP), and audiences (keyword targeting, in-market). Allow 4–6 weeks for PMax learning phase. Use as learning tool, not just marketing channel.
Reference: Marketing Cactus – Using Google Ads to Test Product-Market Fit
Account
├── Campaign: Brand (Search)
├── Campaign: Non-Brand (Search)
├── Campaign: Competitor (Search) — optional; bid on competitor brand + "alternative"/"vs"
├── Campaign: Retargeting (Display)
└── Campaign: Performance MaxWhen: Bid on "Competitor] alternative," "Competitor] vs You]" to intercept high-intent traffic. Google allows competitor terms as keywords; you cannot use competitor names in ad copy without permission.
Landing page: Use a dedicated landing page (comparison/alternatives page), not a blog article. Users searching competitor brands expect direct alternatives—a blog increases bounce; a comparison page matches intent and converts better. See alternatives-page-generator for structure.
Best practices:
Naming: GOOG_[Objective]_[Audience]_[Offer]_[Date] (e.g., GOOG_Search_Brand_Demo_Ongoing)
| Type | Best for | |------|----------| | Search | High-intent queries; keyword-targeted; landing page critical | | Display | Awareness; retargeting; broader reach | | YouTube | Video; awareness; consideration | | Performance Max | Automated; cross-channel; feed + search + display |
Learning period: Run at least 6 weeks for algorithm ramp-up. Works best as complement to Search, not replacement.
Asset groups: Organize by audience intent (e.g., high-intent searchers, cart abandoners, category researchers), not product category alone. Audience signals improve CPA and ROAS vs. no signals.
Asset requirements (per asset group):
Signals: Add remarketing lists and Customer Match to accelerate learning.
Weekly health check: Flag if brand terms >30% of conversions; unexpected geo conversions; any placement >15% of total spend; asset group performance below "Good."
Keyword sources: Use keyword-research for keyword list, clusters, and intent. Map each cluster to a dedicated landing page; relevance improves Quality Score and lowers CPC.
| Factor | Action | |--------|--------| | Expected CTR | Improve ad relevance; test headlines | | Ad relevance | Align ad copy to keyword intent | | Landing page | Ad-to-page alignment; fast load; mobile-friendly |
Target: Quality Score ≥6; higher = lower CPC, better ad rank. Benchmark: Improving Quality Score from 5 to 7 can reduce CPC by 30–50%.
| Conversions/month | Strategy | |-------------------|----------| | <30 | Manual CPC (smart bidding needs volume to optimize) | | 30–50 | Target CPA; minimum for effective smart bidding | | 50–100 | Target CPA | | 100+ | Target ROAS |
Smart bidding: AI-powered bidding (Target CPA, Target ROAS) typically delivers better ROI than manual when conversion volume is sufficient; requires ≥30 conversions in 30 days to work effectively.
When you rank organically (position 4+) for a keyword and also run PPC, paid ads can absorb clicks that would go to organic. Audit: Cross-reference GSC organic rankings with Search Terms report. If organic ranks well, test pausing PPC on those terms to free budget for higher-impact keywords.
Reference: Backlinko – SEO and PPC: 8 Smart Ways to Align
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,180 | 18,139 | -10% | 1 | 1 | 0% | 2,983 | 4,287 | +44% | 0 | 0 | — |
case-02 | fail→pass | 19,724 | 14,288 | -28% | 1 | 1 | 0% | 3,327 | 3,814 | +15% | 0 | 0 | — |
case-03 | pass→pass | 41,614 | 17,850 | -57% | 1 | 1 | 0% | 3,214 | 4,291 | +34% | 0 | 0 | — |
case-04 | fail→pass | 12,005 | 7,040 | -41% | 1 | 1 | 0% | 2,040 | 2,694 | +32% | 0 | 0 | — |
case-05 | pass→pass | 10,624 | 5,170 | -51% | 1 | 1 | 0% | 1,706 | 2,340 | +37% | 0 | 0 | — |
case-06 | fail→pass | 7,197 | 7,266 | +1% | 1 | 1 | 0% | 1,223 | 2,646 | +116% | 0 | 0 | — |
case-12 | pass→pass | 9,393 | 2,486 | -74% | 1 | 1 | 0% | 1,521 | 1,876 | +23% | 0 | 0 | — |
case-07 | pass→pass | 11,785 | 10,103 | -14% | 1 | 1 | 0% | 1,819 | 3,041 | +67% | 0 | 0 | — |
case-08 | fail→pass | 14,206 | 9,397 | -34% | 1 | 1 | 0% | 2,163 | 2,903 | +34% | 0 | 0 | — |
case-09 | fail→pass | 11,479 | 3,019 | -74% | 1 | 1 | 0% | 1,985 | 1,949 | -2% | 0 | 0 | — |
case-10 | fail→fail | 16,817 | 13,785 | -18% | 1 | 1 | 0% | 2,417 | 3,577 | +48% | 0 | 0 | — |
case-11 | fail→fail | 16,111 | 14,785 | -8% | 1 | 1 | 0% | 2,448 | 3,947 | +61% | 0 | 0 | — |
case-13 | fail→pass | 12,450 | 8,857 | -29% | 1 | 1 | 0% | 1,927 | 2,897 | +50% | 0 | 0 | — |
case-14 | fail→pass | 14,833 | 2,559 | -83% | 1 | 1 | 0% | 2,120 | 1,915 | -10% | 0 | 0 | — |
case-15 | pass→pass | 26,024 | 2,760 | -89% | 1 | 1 | 0% | 2,009 | 1,905 | -5% | 0 | 0 | — |
case-16 | fail→pass | 8,625 | 5,438 | -37% | 1 | 1 | 0% | 1,548 | 2,521 | +63% | 0 | 0 | — |
case-17 | pass→pass | 10,288 | 6,067 | -41% | 1 | 1 | 0% | 957 | 2,526 | +164% | 0 | 0 | — |
case-18 | pass→pass | 14,068 | 14,843 | +6% | 1 | 1 | 0% | 2,058 | 2,784 | +35% | 0 | 0 | — |
case-19 | fail→pass | 13,531 | 4,911 | -64% | 1 | 1 | 0% | 2,256 | 2,365 | +5% | 0 | 0 | — |
case-20 | pass→pass | 17,143 | 4,058 | -76% | 1 | 1 | 0% | 1,833 | 2,152 | +17% | 0 | 0 | — |
case-21 | pass→pass | 12,525 | 9,007 | -28% | 1 | 1 | 0% | 2,219 | 3,218 | +45% | 0 | 0 | — |
case-22 | fail→fail | 18,291 | 12,102 | -34% | 1 | 1 | 0% | 2,788 | 3,328 | +19% | 0 | 0 | — |
case-23 | pass→pass | 13,552 | 26,714 | +97% | 1 | 1 | 0% | 2,287 | 3,982 | +74% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +39 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.