ceo-orchestration 1.0.0

package/.claude/skills/domains/paid-media/skills/ppc-strategist/SKILL.md

@@ -0,0 +1,450 @@
+---
+name: ppc-strategist
+description: >
+  Pay-per-click strategy across Google Ads, Microsoft Ads, and Meta: campaign structure
+  design at account/campaign/ad-group/keyword hierarchy, intent-based keyword research
+  with negative-keyword discipline, bid strategy selection matched to data-maturity level,
+  responsive search ad testing with statistical-significance thresholds, landing page
+  message-match and conversion-element validation, budget pacing diagnostics via
+  impression-share signals, and attribution model selection beyond last-click. Use when
+  an account structure lacks tiered isolation between brand, non-brand, competitor, and
+  informational intent; when Smart Bidding is active but conversion-volume prerequisites
+  are unmet; when ad copy tests run without significance thresholds; when budget pacing
+  is managed by gut feel rather than impression-share diagnostics; or when conversion
+  attribution defaults have never been audited against actual customer paths.
+owner: Marcus Holt (PPC Strategist, paid-media domain)
+tier: domain:paid-media
+scope_tags:
+  - ppc
+  - google-ads
+  - search-ads
+  - bid-strategy
+  - ad-copy-testing
+  - attribution
+inspired_by:
+  - source: msitarzewski/agency-agents/paid-media/paid-media-ppc-strategist.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: paid-media
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 500
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/ppc/**"
+  - "**/google-ads/**"
+  - "**/keywords/**"
+  - "**/search-ads/**"
+---
+
+# PPC Strategist
+
+## Cardinal Rule
+
+Account structure is strategy. The architecture of campaigns, ad groups, keyword
+assignments, and bid-strategy containers determines which signals the algorithm
+receives, how budgets flow, and what the advertiser can measure independently. A flat
+account — one campaign, broad match, Smart Bidding, no negatives — is not a simplified
+strategy; it is a strategy that delegates all decisions to the platform without the
+data prerequisites to make those decisions well. Structural choices compound over time:
+a structurally sound account accumulates clean conversion signal, controlled vocabulary,
+and diagnostic clarity; a structurally unsound account accumulates noise that no
+optimisation lever can remove.
+
+## Fail-Fast Rule
+
+Stop and rebuild the account structure when three conditions hold simultaneously:
+(1) Smart Bidding is active on a campaign with fewer than 30 conversions in the last
+30 days at campaign level, (2) there are no campaign-level or account-level negative
+keyword lists, and (3) Search Query Reports contain more than 15% of spend on queries
+that do not map to any intentional keyword theme. Operating bid automation without
+conversion-volume prerequisites accelerates spend on statistically insufficient signals.
+Adding bid optimisation to an account without negatives and without intent segmentation
+does not fix the structural problem — it optimises a broken input faster.
+
+## When to Apply
+
+- A new account buildout requires structural design before any campaign is created.
+- An existing account is being restructured due to efficiency decline, impression-share
+  loss, or uncontrolled CPC growth.
+- Bid strategy is being transitioned from manual CPC to automated bidding and the data
+  prerequisites need verification.
+- Conversion volume has grown sufficiently to reconsider which bid strategy tier is now
+  appropriate.
+- Ad copy tests are running without documented hypotheses, control/variant isolation, or
+  significance thresholds.
+- Budget pacing is uneven, campaigns are hitting daily limits inconsistently, or
+  impression-share lost due to budget exceeds 10%.
+- Attribution defaults have never been reviewed and the account is making budget
+  allocation decisions from last-click data alone.
+- A new platform (Microsoft Ads, Meta) is being added and structural conventions need
+  to be applied consistently.
+
+## Campaign Structure Discipline
+
+Account structure is hierarchical: account → campaign → ad group → keyword/audience.
+Each level serves a distinct governance purpose. Decisions made at the wrong level
+produce either over-segmentation (hundreds of campaigns, each too small for bidding
+signal) or under-segmentation (one campaign absorbing contradictory intent signals that
+no single bid strategy can serve).
+
+**Account-Level Governance**
+
+- Account contains shared negative keyword lists, audience lists, conversion actions,
+  and bidding signals that flow to all campaigns.
+- Conversion action hierarchy is defined at account level: primary conversion actions
+  (those Smart Bidding optimises toward) are separated from secondary conversion actions
+  (those reported but not bid-optimised).
+- Budget ceilings are enforced at campaign level; account-level shared budgets are
+  used only when dynamic budget reallocation across thematically identical campaigns
+  is intentional.
+
+**Campaign-Level Segmentation**
+
+Campaigns are segmented by intent tier, not by keyword theme:
+
+| Campaign Tier | Intent Definition | Bid Strategy Starting Point |
+|---|---|---|
+| Brand | Queries containing the advertiser's own brand terms | Target Impression Share or Manual CPC |
+| Non-Brand | Queries for the product/service category without brand terms | Target CPA or Target ROAS when conversion-volume prerequisites met |
+| Competitor | Queries containing competitor brand terms | Manual CPC or conservative Target CPA; separate budget |
+| Informational | Queries with research or comparison intent, no direct purchase signal | Separate campaign or excluded via negative list depending on LTV model |
+
+Isolating tiers into separate campaigns enables independent budget control, separate
+performance benchmarks, and clean bid-strategy configuration. A single campaign
+combining brand and non-brand queries produces ROAS metrics that are inflated by
+brand-query efficiency and are not meaningful for non-brand spend decisions.
+
+**Ad-Group Structure: SKAG vs STAG vs Broad-Match-Machine-Learning**
+
+Three structural models exist. Each is a deliberate choice, not an arbitrary default:
+
+- **SKAG (Single Keyword Ad Group):** one keyword per ad group; maximum message
+  match between keyword and ad copy; maximum diagnostic isolation. Appropriate when
+  conversion volume per keyword is sufficient for independent tracking and the account
+  operates below the conversion-volume threshold for Smart Bidding at ad-group level.
+  Drawback: maintenance overhead scales linearly with keyword count.
+- **STAG (Single Theme Ad Group):** tightly themed keyword clusters per ad group;
+  shared ad copy reflects the theme rather than the exact keyword; reduces maintenance
+  overhead while preserving thematic relevance. Appropriate for mid-scale accounts
+  where per-keyword isolation is impractical but theme separation is maintained.
+- **Broad-Match with Smart Bidding consolidation:** broad-match keywords feed the
+  algorithm maximum query volume; the algorithm uses audience signals, device context,
+  and conversion history to bid selectively. Requires sufficient conversion volume
+  (minimum 50 conversions per month at campaign level per Google's stated prerequisite),
+  a well-structured negative-keyword list to exclude irrelevant query expansion, and
+  active Search Query Report monitoring. Not appropriate for accounts below the
+  conversion-volume threshold.
+
+**Campaign Type Selection by Intent**
+
+| Campaign Type | When Appropriate | When Not Appropriate |
+|---|---|---|
+| Search | Direct-response; user expresses explicit query intent | Brand awareness without conversion intent; visual-product categories |
+| Shopping / PMax Shopping | E-commerce with product feed; comparison intent | Service businesses; no structured feed |
+| Performance Max | Supplementary reach across Google inventory after Search and Shopping are optimised; sufficient conversion history | Primary campaign type before Search is optimised; insufficient conversion volume |
+| Display | Remarketing to past site visitors; awareness objectives with defined audience | Cold prospecting without remarketing exclusions applied |
+| Demand Gen (Meta / Google) | Upper-funnel creative-driven campaigns with lookalike or interest targeting | Direct-response campaigns where click intent is the primary signal |
+
+## Keyword Research
+
+Keyword research is an intent-classification exercise before it is a volume exercise.
+A keyword with 10,000 monthly searches and mixed intent — some users researching,
+some comparing, some ready to purchase — requires segmentation before it can be bid
+on efficiently. A keyword with 200 monthly searches and unambiguous purchase intent
+may warrant higher bids than the high-volume mixed-intent term.
+
+**Intent Classification**
+
+Every keyword is assigned to one of four intent categories before being added to any
+campaign:
+
+| Intent Class | Signal | Campaign Destination |
+|---|---|---|
+| Brand | Contains the advertiser's trademarked or colloquial brand terms | Brand campaign |
+| Non-Brand Commercial | Describes the product or service without brand reference; user likely in evaluation or purchase phase | Non-brand campaign |
+| Competitor | Contains a competitor's brand or product name | Competitor campaign |
+| Informational | Research or comparison query; no direct purchase signal | Informational campaign if LTV model supports it; otherwise negative list |
+
+**Match-Type Strategy**
+
+Match types control which queries trigger a keyword. Match-type selection is a
+structural decision with bid and attribution consequences:
+
+- Exact match: maximum query control; minimum reach. Use for highest-value, highest-confidence keywords where query-keyword alignment is non-negotiable.
+- Phrase match: moderate control; captures close variants. Use when query expansion around a core phrase is acceptable and monitored.
+- Broad match: maximum reach; minimum control. Use only when Smart Bidding is active, conversion-volume prerequisites are met, and a comprehensive negative-keyword list is in place.
+
+Mixed match types within the same ad group for the same keyword intent produce
+auction cannibalism — exact and phrase variants compete against each other and
+produce ambiguous Quality Score signals.
+
+**Negative-Keyword Discipline**
+
+Negative keywords are not optional hygiene; they are structural controls that define
+the account's query scope. An account without negative keywords is bidding on the
+full query space the platform chooses to match, which includes irrelevant, low-intent,
+and competitive queries that consume budget without conversion signal.
+
+- Campaign-level negatives: intent-class exclusions. The non-brand campaign excludes
+  all brand terms; the brand campaign excludes all non-brand commercial terms.
+- Ad-group-level negatives: theme exclusions within a campaign. An ad group for
+  "project management software" excludes "free" if the product has no free tier.
+- Account-level shared negative lists: universal exclusions (competitor names in
+  non-competitor campaigns, irrelevant industry terms, navigational queries) applied
+  across all campaigns.
+- Search Query Report review cadence: weekly for active accounts; negatives added
+  within 7 days of identifying an irrelevant spend pattern.
+
+## Bid Strategy Selection
+
+Bid strategy is matched to data maturity. The correct bid strategy for an account
+with 10 conversions per month is not the same as the correct strategy for an account
+with 500 conversions per month. Applying Smart Bidding before the data prerequisites
+are met does not accelerate learning; it produces auction decisions based on
+statistically insufficient conversion signals.
+
+**Bid Strategy Matrix by Data Maturity**
+
+| Strategy | Conversion-Volume Prerequisite | When to Select | Risk Profile |
+|---|---|---|---|
+| Manual CPC | None — appropriate below any automated threshold | New accounts; brand campaigns requiring impression-share control; accounts where conversion tracking is not yet verified | Full manual exposure to auction volatility; requires active daily management |
+| Enhanced CPC (eCPC) | Minimum 20–30 conversions per month | Transition layer from manual to automation; retains manual bid floor with algorithmic micro-adjustments | Low automation risk; limited upside vs Target CPA |
+| Maximise Conversions | Minimum 30 conversions per month at campaign level; no CPA target | Budget-constrained campaigns where spend efficiency is secondary to volume growth; time-limited promotions | Budget exhaustion risk without CPA guardrail |
+| Target CPA | Minimum 30–50 conversions per month at campaign level | Established campaigns with stable conversion rate; known acceptable cost-per-acquisition | CPA target set too low starves the campaign of impression share; target set too high produces budget waste |
+| Target ROAS | Minimum 50 conversions per month at campaign level; revenue values attached to conversions | E-commerce accounts with heterogeneous product values; portfolio bid strategies across multiple campaigns | Revenue-value accuracy dependency; incorrect values produce incorrect bidding |
+| Maximise Conversion Value | Same as Target ROAS without explicit ROAS target | Revenue maximisation when a specific ROAS target cannot yet be set | Same revenue-value accuracy dependency as Target ROAS |
+
+**Bid Strategy Transition Protocol**
+
+Transitions between bid strategies require a transition window. Abrupt strategy
+changes reset the algorithm's learning period and can produce 2–4 weeks of
+performance instability:
+
+1. Verify conversion-volume prerequisites are met for the target strategy.
+2. Set a conservative initial target (CPA target 20% above current actual CPA; ROAS
+   target 20% below current actual ROAS) to allow the algorithm to maintain impression
+   share during the learning period.
+3. Do not change budgets, add/remove keywords, or modify audiences during the first
+   14 days of a new bid strategy.
+4. Evaluate performance at day 14 and day 30 before adjusting targets.
+
+## Ad Copy Testing
+
+Ad copy testing is a structured experiment, not a continuous rotation. An account
+running three headline variants in a Responsive Search Ad (RSA) without a documented
+hypothesis, a control definition, and a significance threshold is not running a test;
+it is running an uncontrolled exposure that produces inconclusive data.
+
+**Responsive Search Ad Configuration**
+
+RSAs allow up to 15 headlines and 4 descriptions. Google's algorithm tests
+combinations and reports an "Ad Strength" score. Ad Strength is a proxy metric — it
+measures variation coverage, not conversion performance.
+
+- Pin headlines 1 and 2 for brand-consistency and message-match elements that must
+  always appear. Pinning reduces the algorithm's combination freedom; use pinning
+  only for elements where variant exposure is unacceptable.
+- Leave headline positions 3–15 and all description positions unpinned for
+  algorithmic combination testing.
+- Include at least one headline variant that directly reflects the most common
+  query intent for the ad group. Message match between query and headline is a
+  Quality Score input.
+
+**Structured A/B Testing Protocol (RSA vs RSA)**
+
+When testing fundamentally different value propositions — not headline variants within
+the same RSA — use ad variation experiments or separate ad group A/B tests:
+
+1. Define one hypothesis: "Replacing a feature-benefit headline with a social-proof
+   headline will increase conversion rate by X%."
+2. Define the control (current top-performing RSA) and the variant (one substantive
+   change from the control).
+3. Run both ads simultaneously at 50/50 rotation.
+4. Determine the minimum sample size required for statistical significance at 95%
+   confidence before the test starts. A test declared complete before reaching
+   minimum sample size produces a false conclusion.
+5. Do not modify campaign settings, budgets, or bid strategies during the test.
+6. Declare a winner only after reaching significance. If significance is not reached
+   within 60 days, the test is inconclusive — do not declare the variant a winner
+   based on directional trends.
+
+**Headline and Description Rotation Rules**
+
+- Rotate by conversions (not clicks) when conversion volume is sufficient (≥30
+  per variant per month).
+- Rotate evenly when conversion volume is insufficient to declare conversion-rate
+  significance; use click-through rate as a directional signal only.
+- Document each active test with hypothesis, start date, sample size target, and
+  current status. Undocumented tests are invisible to account transitions.
+
+## Landing Page Conformance
+
+The click is not the outcome. A campaign with a 5% click-through rate and a 0.5%
+conversion rate is not a campaign with a good CTR — it is a campaign with a broken
+post-click path. Landing page conformance is a prerequisite audit, not a post-launch
+optimisation.
+
+**Message Match**
+
+Message match is the alignment between the query, the ad headline, and the landing
+page headline. A user searching for "enterprise invoicing software" who clicks an ad
+headlined "Enterprise Invoicing Software" must arrive at a page headlined around the
+same concept. Each break in the message-match chain is a conversion-rate friction point.
+
+- Verify message match for every ad group by tracing: primary keyword → ad headline
+  → landing page headline → primary CTA. Gaps at any link require correction before
+  campaign launch.
+
+**Above-the-Fold Conversion Elements**
+
+A landing page that requires scrolling to find the primary conversion element (form,
+CTA button, phone number, product add-to-cart) is consuming the attention threshold
+that the ad spend purchased. Required above-the-fold elements:
+
+- Primary headline aligned to ad message.
+- One clear CTA with unambiguous action language.
+- At minimum one trust signal (social proof, security badge, or guarantee statement).
+
+**Load-Time Budget**
+
+A landing page that loads in more than 3 seconds on a mobile connection loses a
+disproportionate share of mobile-click conversions. Mobile-first is not an optimisation
+preference; it is a budget-protection measure on accounts where more than 40% of
+clicks originate from mobile devices.
+
+- Verify Core Web Vitals (LCP, INP, CLS) for every landing page before campaign
+  launch and after any page modification.
+- A campaign driving paid traffic to a page with LCP above 4 seconds is paying for
+  clicks that the page design is discarding.
+
+## Budget Pacing
+
+Budget pacing is not synonymous with budget spending. A campaign that spends 100%
+of its daily budget at noon has not paced correctly — it has exhausted its reach
+window for the remainder of the day. Pacing discipline matches spend velocity to the
+audience's conversion-probability distribution across hours and days.
+
+**Even Pacing vs Front-Load Pacing**
+
+| Pacing Model | When to Apply | Risk |
+|---|---|---|
+| Even pacing (default) | Conversion probability is relatively uniform across hours; no strong intraday conversion-rate pattern | Under-serves peak-conversion hours if the distribution is not actually uniform |
+| Front-load pacing | Conversion rate is materially higher in the first half of the day; time-sensitive offers; auction competition is lower in morning hours | Exhausts budget before peak-search-volume periods if analysis is incorrect |
+
+**Impression Share Lost Diagnostics**
+
+Impression Share (IS) reports distinguish between two loss types: IS lost due to
+budget and IS lost due to rank. These have different corrective actions:
+
+- IS lost to budget: the campaign is eligible to show but has no remaining budget.
+  Corrective actions: increase daily budget, or improve Quality Score to reduce CPC
+  and extend budget reach.
+- IS lost to rank: the campaign is losing auctions to competitors with higher Ad Rank.
+  Corrective actions: improve Quality Score (expected CTR, ad relevance, landing page
+  experience) or increase bids. Increasing budget does not correct IS lost to rank.
+
+Diagnosing IS loss incorrectly and applying the wrong corrective action (increasing
+budget for IS lost to rank, or increasing bids for IS lost to budget) is a structural
+mismanagement pattern.
+
+**Spend-Velocity Alerts**
+
+- Set daily spend alerts at 80% of daily budget by noon local time. Consistent
+  early exhaustion is a structural signal requiring pacing model review.
+- Set underpacing alerts at 30% of daily budget remaining by 5 PM local time.
+  Consistent underpacing indicates Quality Score, bid, or targeting constraints that
+  are restricting impression eligibility below budget capacity.
+
+## Attribution Discipline
+
+Default attribution models misrepresent the contribution of campaigns to conversions.
+A paid search account relying exclusively on last-click attribution systematically
+under-values non-brand campaigns (which are often first-touch or assist-touch), and
+systematically over-values brand campaigns (which are often last-touch on traffic
+initiated by non-brand discovery). Attribution model selection is a budget allocation
+decision disguised as a measurement setting.
+
+**Attribution Model Comparison**
+
+| Model | What It Measures | Best Use |
+|---|---|---|
+| Last click | Final click before conversion | Simple baseline; directional only; systematically biases toward brand and retargeting |
+| First click | First paid click in the path | Upper-funnel contribution measurement; not a complete picture |
+| Linear | Equal credit across all clicks in path | Multi-touch awareness; no differential weighting |
+| Time decay | More credit to clicks closer to conversion | Appropriate when recency is a genuine purchase driver |
+| Data-driven (DDA) | Algorithmic credit assignment based on observed path patterns | Preferred for accounts with sufficient conversion volume (≥300 conversions per month); most accurate for bidding alignment |
+
+**Conversion Window Selection**
+
+Conversion windows define the period after a click during which a subsequent
+conversion is attributed to that click. Default windows (30 days for purchases, 90
+days for lead forms in Google Ads) do not always match the actual purchase cycle:
+
+- B2C impulse products: conversion window 1–7 days.
+- B2C considered purchases: conversion window 14–30 days.
+- B2B lead-to-close cycles: conversion window 30–90 days; view-through attribution
+  requires separate justification and is not included in primary bid-optimisation
+  conversion actions.
+
+Setting a 30-day conversion window for a product with a 2-day average purchase
+cycle inflates attributed conversions and overstates campaign contribution.
+
+**Lift Studies and Incrementality**
+
+Attribution models measure credit allocation; they do not measure incrementality.
+A campaign that receives credit for 500 conversions in a last-click model does not
+prove that those 500 conversions would not have occurred without the campaign. Lift
+studies (geo-holdout, matched-market, or platform-native incrementality tests) are
+the correct tool for measuring whether a campaign is generating incremental demand
+or capturing demand that would have converted through another path.
+
+Accounts making budget-scaling decisions from attribution data alone, without an
+incrementality frame, risk scaling campaigns that are capturing existing intent rather
+than generating new demand.
+
+## Anti-patterns
+
+| Anti-pattern | Failure Mode |
+|---|---|
+| Spray-and-pray match types | All keywords set to broad match without negative-keyword lists or conversion-volume prerequisites for Smart Bidding; the account bids on the full query space the platform chooses to match, including irrelevant, competitor-adjacent, and informational queries that consume budget without commercial-intent signal |
+| No negatives | An account operating with no campaign-level or account-level negative keywords has an undefined query scope; every broad or phrase-match keyword is bidding on a superset of intended queries; Quality Score, conversion rate, and average CPC all degrade from irrelevant traffic contamination |
+| Vanity-CTR optimisation | Optimising for click-through rate as a primary KPI produces high-CTR, low-intent traffic; CTR is a Quality Score input and a relevance signal, not a conversion predictor; accounts managed to CTR targets routinely achieve high impressions and low conversion volume |
+| Ignoring Search Query Report | The Search Query Report (SQR) is the account's ground truth for what queries are actually triggering ads; an account that does not review the SQR weekly has no visibility into the query-to-keyword gap; negatives are not added; irrelevant spend accumulates silently |
+| No landing-page test | Campaign optimisation without landing page experimentation treats the post-click path as a constant; in most accounts, landing page conversion rate variance across variants exceeds bid strategy efficiency gains; optimising the ad without testing the destination is optimising the top of the conversion path while ignoring the bottom |
+| Smart Bidding below prerequisite threshold | Deploying Target CPA or Target ROAS on campaigns with fewer than 30 conversions per month per the stated minimum; the algorithm lacks sufficient signal to bid selectively; the result is either aggressive spend on low-signal queries or excessive bid suppression from premature learning-phase signals |
+| Conflating attribution credit with incrementality | Treating high attributed conversion volume as evidence that a campaign is generating demand; attribution models redistribute credit, they do not measure whether the credited conversions are incremental; scaling a campaign that captures existing intent rather than generating new demand inflates spend without proportional revenue lift |
+
+## Cross-References
+
+- `domains/paid-media/skills/search-query-analyst` — Search Query Report analysis,
+  query-to-keyword gap identification, and negative-keyword expansion protocols.
+- `domains/paid-media/skills/auditor` — Full paid-media account audit framework
+  covering structural health, bid strategy compliance, conversion tracking verification,
+  and attribution model review.
+- `domains/paid-media/skills/tracking-specialist` — Conversion tracking implementation,
+  tag audit, Google Tag Manager configuration, and conversion-action hierarchy setup.
+
+## ADR Anchors
+
+- **ADR-058** (Brainstorm Gate and Two-Pass Review): PPC strategy authorship follows
+  the two-pass discipline — pass one is structural analysis and diagnostic (account
+  audit, intent classification, bid-strategy data-maturity check, attribution model
+  review) without drafting any campaign changes; pass two is campaign design and
+  optimisation recommendations using only what the diagnostic surfaced. A campaign
+  restructure authored before the structural audit is complete produces a structure
+  optimised for the strategist's preferences rather than the account's actual conversion
+  signal, the paid-media equivalent of a reviewer who generates and evaluates in a
+  single motion.