Google Ads

Google Ads — Pro Playbook

A practitioner's guide to architecting, launching, and scaling Google Ads accounts. Use real platform terminology and current features. When the user has connected the bundled google-ads MCP (uvx googleads/google-ads-mcp), drive it for live account operations: read with gads_google_ads_search_google_ads (GAQL), build with the gads_campaign_*, gads_ad_group_*, gads_asset_group_*, gads_bidding_strategy_* and gads_keyword_* tools, and pull keyword ideas with gads_keyword_plan_idea_*. Always read before you write, and stage large changes through gads_batch_job_*.

> Reuse note: this skill is authored fresh for SquadCoder; structure informed by public MIT Google Ads > skill packs (e.g. TheMattBerman/google-ads-copilot, nowork-studio/NotFair).

---

1. Account & campaign structure

Hierarchy: MCC (manager) → Account → Campaign → Ad group / Asset group → Ads / Assets → Keywords / Audiences.

Pick the campaign type by intent and inventory:

• Search — high-intent text ads on the SERP. The backbone of most accounts. Use intent-tight ad groups.
• Performance Max (PMax) — one campaign serving all Google inventory (Search, Shopping, Display, YouTube,

Discover, Gmail, Maps) via asset groups + audience signals. Goal-based; great for ecommerce + lead-gen once you have conversion data. It cannibalizes Shopping/brand Search, so guard brand terms.

• Shopping — product ads from a Merchant Center feed. Standard Shopping gives priority/placement control;

PMax replaces "Smart Shopping". Use feed quality (titles, GTINs, product types) as your lever.

• Demand Gen — visually rich social-style ads on YouTube (in-feed, Shorts), Discover, and Gmail. Use for

mid-funnel demand creation with lookalike ("similar") segments and your 1st-party lists.

• Display — GDN banners/responsive display. Cheap reach + remarketing; weak for cold prospecting alone.
• YouTube / Video — objectives: brand awareness/reach, consideration (in-stream skippable, in-feed),

action (Video action campaigns / VAC). Use bumpers (6s) for frequency, skippable in-stream for storytelling.

Account architecture principles:

• One theme per ad group. Tight ad group → relevant RSA → high Quality Score → lower CPC.
• Separate brand vs non-brand campaigns (different intent, CPC, and budgets; protects brand ROAS reporting).
• Isolate competitor and high-spend terms so budget and negatives don't bleed across themes.
• STAG/SKAG is mostly dead post-close-variants — prefer tightly-themed ad groups over single-keyword groups.
• Use shared budgets sparingly; prefer per-campaign budgets for clean control unless portfolio-bidding.

---

2. Keywords, match types & negatives

Match types (current behavior):

• Broad match — widest reach; relies on Smart Bidding + audience signals to stay efficient. Only use broad

with a conversion-based Smart Bidding strategy and a solid negative list. Never broad on manual CPC.

• Phrase match "..." — meaning of the phrase, order respected; the workhorse for controlled scale.
• Exact match [...] — same meaning/intent; tightest control, best for proven converters.

Strategy: start phrase + exact on proven terms; layer in a broad-match + tCPA test campaign to mine new queries, then harvest winning search terms into phrase/exact and add the rest as negatives.

Negatives are non-negotiable:

• Build an account-level negative keyword list (shared set) for junk: "free", "jobs", "salary", "תמונות",

"חינם", "משרות", plus irrelevant brands.

• Mine the Search Terms report weekly; add irrelevant terms as negatives at the right level.
• Use negatives to sculpt traffic between brand/non-brand and between ad groups.
• Mind match type on negatives (negative broad does NOT match close variants/synonyms the way positive broad does).

---

3. Bidding & budget strategy

Smart Bidding (auction-time, conversion-based) — pick by goal:

• Maximize Conversions — spend the budget to get the most conversions. Use at launch when learning volume;

add an optional tCPA cap once you have ~15–30 conversions/30 days.

• Target CPA (tCPA) — hit an average cost per conversion. Lead-gen default once stable.
• Maximize Conversion Value — most value within budget; the value-based launch mode for ecommerce.
• Target ROAS (tROAS) — hit a revenue/spend ratio. Requires conversion values; the ecommerce endgame.

Set tROAS slightly below current achieved ROAS to avoid throttling, then ratchet up.

• Target Impression Share — for brand defense / absolute-top placement; not a performance strategy.
• Manual CPC / Enhanced CPC — legacy; avoid except niche control cases. eCPC is being deprecated.

Portfolio bid strategies let several campaigns share one tCPA/tROAS target — useful for thin-data campaigns.

Budget pacing & CBO/ABO equivalent:

• Google has no CBO toggle like Meta. Campaign-level budget = ABO (the norm); **portfolio strategy +

shared budget ≈ CBO** behavior. Default to per-campaign budgets for control.

• Daily budget can spend up to 2× on a given day, capped at the monthly average — don't panic at daily spikes.
• Give Smart Bidding ~7–14 days to exit the learning period after major changes; avoid budget/target

swings >20% at once. Use seasonality adjustments for short spikes (sale weekend), data exclusions to mask broken-tracking days.

---

4. Creative specs & best practices

Responsive Search Ads (RSA) — the only Search ad unit:

• Up to 15 headlines (30 char) + 4 descriptions (90 char). Aim for ≥8–10 distinct headlines.
• Include the keyword/theme in 2–3 headlines; pin sparingly (pin 1–2 must-have headlines to positions, leave

the rest unpinned so Google can mix). Over-pinning kills Ad Strength.

• Target "Good"/"Excellent" Ad Strength; add benefits, offers, CTAs, social proof, and a DKI/customizer

where relevant. Use ad customizers + countdown for dynamic pricing/urgency.

• Assets (formerly extensions): always run sitelinks (≥4), callouts (≥4), structured snippets, plus

call, lead form, image, price, promotion, and location where applicable. Assets lift CTR for free.

Performance Max asset groups: per group supply text (headlines/long headlines/descriptions), images (1.91:1, 1:1, 4:5 — min 600×314 / 300×300), logos (1:1, 4:1), and video (≥10s; supply your own or Google auto-generates — always upload real video). Add audience signals (your data + custom segments) to speed learning.

Display / Demand Gen: responsive display assets (landscape 1.91:1, square 1:1, portrait 4:5/9:16 for Demand Gen + Shorts). Lead with the hook in the first 2s; design for sound-off and mobile.

YouTube: skippable in-stream needs a hook before the 5s skip; bumpers (6s, non-skippable) for reach/frequency; use a clear end-card CTA. Vertical/Shorts assets for Demand Gen and VAC.

Shopping: the "creative" is your feed — optimize titles (brand + product + key attributes first 40 chars), high-res images, correct GTIN/MPN, product_type/Google category, and competitive price.

---

5. Conversion tracking & measurement

This is the foundation — Smart Bidding is only as good as the conversion data.

• Google tag (gtag.js) or Google Tag Manager sitewide; define conversion actions in Google Ads

(or import from GA4). Mark only true business outcomes as Primary; demote micro-events to Secondary.

• Enhanced Conversions — send hashed (SHA-256) 1st-party data (email/phone) to recover conversions lost to

cookie/consent gaps. Enable Enhanced Conversions for web (and for leads in lead-gen) — meaningful uplift.

• Server-side / offline: upload offline click conversions (GCLID) for CRM-qualified leads/closed deals so

bidding optimizes to revenue, not raw form fills. Use conversion value rules to adjust value by geo/device/audience.

• Consent Mode v2 (update_default/update) is required for EEA traffic and feeds conversion modeling.
• Attribution: Google Ads now uses data-driven attribution (DDA) by default — last-click is deprecated.
• Via MCP: create actions with gads_conversion_create_conversion_action, push Enhanced Conversions with

gads_user_data_upload_enhanced_conversions, and upload offline conversions with gads_conversion_upload_upload_click_conversions.

---

6. Optimization workflow (cadence & rules)

Daily (5 min): check for disapprovals, billing/budget-limited flags, tracking breakage (0 conversions where there should be some), and sudden CPA/ROAS swings.

2–3× / week:

• Search Terms report → harvest winners (add as exact/phrase), add junk as negatives.
• Watch campaigns in learning after edits; don't stack changes.
• Review asset/RSA Ad Strength and swap low-performing assets.

Weekly:

• Bid-target tuning: nudge tCPA/tROAS by ≤10–15% toward goal; never yank targets.
• Pause or restructure ad groups/keywords with spend but no conversions over a statistically meaningful window.
• Check auction insights (impression share lost to rank vs budget) to decide raise-bid vs raise-budget.
• Review Recommendations tab — apply the good ones (negatives, new keywords) skeptically; decline auto-apply

of broad-match/budget recs unless intended.

Scaling rules: when a campaign hits target and is limited by budget (impression share lost to budget > 10%), raise budget 15–25% at a time, then let it re-stabilize. To scale efficiency, loosen tROAS / raise tCPA in small steps to unlock volume. Use Experiments / drafts (gads_experiment_*) to A/B structural or bidding changes before rolling out.

Kill rules: pause keywords/ad groups/products with clearly-above-target CPA and enough data; pause placements (Display/Video) with spend + zero conversions; exclude irrelevant or brand-unsafe placements and content labels.

---

7. Common mistakes / pitfalls

• Broad match on manual CPC with no Smart Bidding and weak negatives → budget incineration.
• No / wrong conversion tracking (counting every page view, double-counting, or tracking broken) → bidding to noise.
• Too many primary conversion actions of unequal value → the algorithm optimizes to the cheap one.
• Over-pinning RSA headlines → "Poor" Ad Strength, low reach.
• Letting PMax cannibalize brand → add brand as account negatives to PMax (via support/exclusions) and run a

dedicated brand Search campaign.

• Reacting to single days / changing targets daily → perpetual learning, never stabilizes.
• Ignoring the Search Terms & placement reports → wasted spend on junk queries and MFA sites.
• Geo set to "presence + interest" by accident → ads show to people merely searching about your area. Use

"Presence: people in your targeted locations" for local businesses.

• Auto-applied recommendations left on → silent broad-match/budget changes you didn't intend.

---

8. Israel / Hebrew market notes

• Language & targeting: target Hebrew (and often Arabic + Russian for relevant segments) plus

English for tech/B2B. Set Location = Israel (or specific cities) with "presence" targeting for local.

• RTL creative: write Hebrew headlines/descriptions natively RTL — verify line breaks, punctuation (?/!/,

placement), and that numbers/Latin brand names render correctly inside RTL text. Don't machine-translate ad copy; Hebrew CTAs differ ("קבל הצעת מחיר", "השאירו פרטים", "קנו עכשיו").

• Hebrew keyword nuance: Hebrew is written without niqqud and has heavy morphology (prefixes ב/ל/ה/ו/מ/ש,

plural/gender forms, construct state). Cover spelling variants (with/without כתיב מלא extra yod/vav), common transliterations of brand/English terms, and slang. Lean on broad + Smart Bidding to catch variants, then harvest. Search volumes are smaller than EN — don't over-fragment ad groups.

• Budgets in NIS (₪): quote and pace budgets in shekels; CPCs vary widely by vertical (legal/insurance/real-estate

are expensive). Account currency is set once at creation — confirm before building.

• Shabbat & holidays: Israeli buying patterns dip Friday afternoon–Saturday and around חגים (Rosh Hashana,

Yom Kippur, Sukkot, Pesach). Use ad scheduling + seasonality adjustments; for some audiences shift weight to Sunday–Thursday. Plan promo pushes around חגים and local sale events.

• Privacy/consent: Israel follows its own privacy law (Protection of Privacy Law) plus growing GDPR-style

expectations; implement Consent Mode and clear consent UX, especially for EU-facing traffic.

---

9. Output (what to produce when this skill activates)

When asked to plan or build, produce a concrete, ready-to-implement deliverable:

1. Campaign skeleton — campaign type(s), naming, budget (₪/day), bid strategy + target, locations, languages,

ad schedule; brand/non-brand split.

1. Ad group / asset group plan — themes with keywords by match type and seed negative keyword list

(account + campaign level); for PMax/Shopping, the audience signals and feed/asset-group breakdown.

1. Creative — at least 3 RSA variants (≥10 headlines + 4 descriptions each, Hebrew + English where relevant),

plus the full asset/extension set (sitelinks, callouts, snippets, images, etc.).

1. Tracking plan — conversion actions (primary/secondary + values), Google tag/GTM + Enhanced Conversions,

offline-conversion plan for lead-gen, DDA attribution.

1. Optimization & scaling plan — review cadence, target-tuning steps, scaling and kill rules, experiment plan.

If the google-ads MCP is connected, offer to apply the plan live (create budget → campaign → ad groups → keywords → RSAs → conversion actions), reading current state first and staging bulk writes via a batch job. Always confirm spend-affecting changes with the user before executing.

Source & license

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

• Author: squadcodercom
• Source: squadcodercom/squadcoder
• License: MIT
• Homepage: https://squadcoder.com

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