Pattern: Parallel Workers | Team size: 6 agents
This team designs multiple optimization experiments in parallel (bidding, creatives, audiences, landing pages) to accelerate learning and performance gains for Pasyros. Parallel workers fit because these workstreams can be developed and validated independently, then rolled into a unified test calendar.
google ads pasyros analitikas
Create an agent team to design a rapid, parallel Google Ads optimization experiment program for “Pasyros” (project: google_ads_pasyros_analitikas). Your mission is to generate a ready-to-execute 4-week experimentation plan with clear hypotheses, setup steps, measurement, decision rules, and a unified weekly calendar. Assume you are advising an in-house marketing team running Google Ads (Search, Performance Max, remarketing). If any critical context is missing (industry, conversion, geo, budget), make explicit assumptions and list exactly what must be confirmed.
GLOBAL REQUIREMENTS
- Language: Lithuanian, but keep platform terms (tCPA, tROAS, RSA, PMax, GA4) in English.
- Deliverables must be practical and copy-paste-ready: tables, checklists, naming conventions.
- Every agent must challenge assumptions: include a “Rizikos / prielaidos” section with at least 5 bullets and how to mitigate/validate.
- Dependencies must be explicit: if a step requires another deliverable, state “Šis darbas turi būti baigtas prieš …”.
- Output files: write ONLY to these paths:
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/00_context_assumptions.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/01_test_program_lead.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/02_bidding_budget_experimenter.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/03_creative_variant_builder.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/04_audience_remarketing_strategist.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/05_landing_page_cro_analyst.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/06_measurement_readout_analyst.md
- outputs/agent_teams_demo/google_ads_pasyros_analitikas/07_synthesis_master_plan.md
- Word counts (hard targets, ±10%):
- 00_context_assumptions.md: 350–450 words
- Each agent file (01–06): 900–1,200 words
- 07_synthesis_master_plan.md: 1,100–1,400 words
- Collaboration mechanics:
- Use a shared “Insights Exchange” protocol: each agent must provide 8–12 bullet “Įžvalgos kitiems agentams” and 5–8 bullet “Klausimai / kas man trūksta” at the end of their file.
- Each agent must explicitly reference at least 2 ideas from other agents’ outputs in their own plan (after those files exist). If working concurrently, add a placeholder “(laukiu X agento įžvalgų)” then update in synthesis.
- Each agent must include “Kas gali nepavykti?” with at least 5 failure modes and detection signals.
- Testing rigor:
- Each proposed experiment must include: hypothesis, setup steps, targeting/scope, KPI primary/secondary, guardrails, expected direction, minimum runtime, stop/continue rules, and what to do if results are inconclusive.
- Prefer incremental, low-risk tests first, then bolder tests.
PROJECT CONTEXT BOOTSTRAP (MUST COMPLETE FIRST)
Task: Create baseline assumptions and a minimal briefing checklist so the rest of the team can work without waiting.
- Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/00_context_assumptions.md with:
1) “Trumpas aprašymas” (2–3 sentences) of what “google ads pasyros analitikas” likely means.
2) “Prielaidos (kol nepatvirtinta)” table with 10 rows: Industry, offer type, conversion definition, geo, language, budget range, current channels, current tracking stack, sales cycle length, key constraints. Put assumed values + “Reikia patvirtinti: taip/ne”.
3) “Ką reikia gauti iš komandos per 48 val.” checklist of 12 items (e.g., conversion goals, CRM stages, margins, feed availability, creative assets, consent mode, etc.).
4) “Bendros guardrails” (6 bullets): e.g., don’t break tracking, don’t increase CPA >X% during learning, protect brand terms, etc.
Dependency: 00_context_assumptions.md must be completed before any other agent finalizes their recommendations (others can draft but must reconcile with 00).
AGENT 1 — TEST PROGRAM LEAD
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/01_test_program_lead.md.
Responsibilities: define overall hypothesis backlog, prioritization, success metrics, guardrails, and a weekly experiment calendar that coordinates all parallel workstreams.
Deliverable structure (use these exact headings):
1) “Tikslas ir šiaurės žvaigždė” (define a single North Star metric + 3 supporting metrics)
2) “Eksperimentų principai ir guardrails” (10 bullets, including budget reallocation limits, brand protection, learning periods)
3) “Hipotezių backlog’as (prioritetuotas)” table with 12–18 experiments (rows) and columns: ID, Sritis (Bidding/Budget, Creative, Audience, LP CRO, Measurement), Hipotezė, Poveikis (H/M/L), Pastangos (H/M/L), Rizika (H/M/L), Priklausomybės, Reikalingi asset’ai, Pagrindinis KPI, Guardrail KPI, Siūloma trukmė, Owner (agent)
4) “ICE/PIE vertinimas” (explain scoring in 120–180 words) + a scoring table for top 8 experiments with numeric scores
5) “4 savaičių eksperimentų kalendorius” (weekly table W1–W4) including: launches, monitoring checkpoints, decision dates, and which experiments must not overlap
6) “Sprendimų taisyklės (stop/scale/iterate)” (a decision tree in text + explicit thresholds; include inconclusive handling)
7) “Rizikos / prielaidos” (≥5) + mitigations
8) “Kas gali nepavykti?” (≥5) + detection signals
9) “Įžvalgos kitiems agentams” (8–12 bullets)
10) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must incorporate measurement constraints from Agent 6 before finalizing decision rules.
- Must incorporate overlap/conflict notes from Agents 2–5 before finalizing the calendar.
Collaboration: challenge at least 2 assumptions from 00_context_assumptions.md and propose validation steps.
AGENT 2 — BIDDING & BUDGET EXPERIMENTER
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/02_bidding_budget_experimenter.md.
Responsibilities: propose and simulate changes to bidding strategies, tROAS/tCPA targets, campaign structure budget allocation, and query controls to improve efficiency.
Deliverable structure:
1) “Dabartinės būklės diagnostika (prielaidos)” (what you’d inspect: 12-point checklist)
2) “Bidding eksperimentai” — propose 4–6 experiments with full templates:
- Experiment name, Hypothesis, Where (Search/PMax), Setup (steps in Google Ads UI), Target setting (tCPA/tROAS values with rationale), Budget change (% and cap), Exclusions/controls, KPI, Guardrails, Runtime, Stop/Scale rules
3) “Budget allocation plan (draft)” table: Campaign type, Current share (assumed), Proposed share, Rationale, Risk, Dependency
4) “Query & brand protection” (specific negative keyword and brand separation recommendations; include a naming convention)
5) “Learning period and volatility management” (8 bullets: how to avoid resets, when to avoid changes)
6) “Rizikos / prielaidos” (≥5) + mitigations
7) “Kas gali nepavykti?” (≥5) + detection signals
8) “Įžvalgos kitiems agentams” (8–12 bullets)
9) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must align KPI definitions and conversion mapping with Agent 6 measurement plan before final tCPA/tROAS recommendations.
- Must check with Agent 1 calendar to avoid overlapping major bid strategy changes with landing page tests in the same traffic slice.
AGENT 3 — CREATIVE VARIANT BUILDER
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/03_creative_variant_builder.md.
Responsibilities: generate ad copy and asset variations aligned to top intents and offers, ready for RSA and PMax testing.
Deliverable structure:
1) “Intentų žemėlapis” table with 6–10 intent themes (e.g., brand, price, problem, competitor, how-to, local, urgent) and recommended campaign/ad group placement
2) “RSA paketas”:
- Provide 6 RSA “ad sets”. Each ad set includes:
- 12 headlines (≤30 chars each; label: H1–H12)
- 4 descriptions (≤90 chars each; label: D1–D4)
- 2 path suggestions
- Pinning guidance (what to pin and why; include at least one no-pin variant)
- Include 1 brand-safe and 1 aggressive offer-driven set.
3) “PMax asset kit”:
- 8 short headlines, 6 long headlines, 6 descriptions
- 3 value propositions, 6 audience signals (seed ideas), 6 image concepts (no actual images), 2 video storyboard concepts (15–20 sec)
4) “Creative experiments” (3–5) with hypotheses and measurement: what to test, which assets swap, success metric, guardrail, runtime
5) “Ad policy & compliance checklist” (10 items)
6) “Rizikos / prielaidos” (≥5) + mitigations
7) “Kas gali nepavykti?” (≥5) + detection signals
8) “Įžvalgos kitiems agentams” (8–12 bullets)
9) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must coordinate with Agent 4 on audience intent/segment messaging.
- Must coordinate with Agent 5 on landing page promise alignment (message match) before finalizing offer-heavy claims.
AGENT 4 — AUDIENCE & REMARKETING STRATEGIST
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/04_audience_remarketing_strategist.md.
Responsibilities: design audience layering, remarketing segments, exclusions, and traffic quality controls to raise conversion rate and reduce low-quality traffic.
Deliverable structure:
1) “Segmentavimo schema” diagram-in-text + table defining 8–12 segments: Source, Membership duration, Inclusion criteria, Exclusions, Use case (Observation/Targeting/PMax signals), Risk
2) “Remarketing programos dizainas”:
- 3 remarketing funnels (short, mid, long cycle) with recommended durations, creatives, offers, and frequency/recency guidance
3) “Exclusions & quality controls”:
- Lists for: existing customers, converters, job seekers, irrelevant geos/languages, low-intent placements (as applicable)
4) “Audience experiments” (4–6) with full templates: hypothesis, setup, scope, KPI, guardrails, runtime, stop/scale
5) “Privacy & consent considerations” (GA4, Consent Mode, enhanced conversions: what to verify; 8 bullets)
6) “Rizikos / prielaidos” (≥5) + mitigations
7) “Kas gali nepavykti?” (≥5) + detection signals
8) “Įžvalgos kitiems agentams” (8–12 bullets)
9) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must align with Agent 6 on audience definition feasibility in GA4/Google Ads and attribution windows.
- Must align with Agent 2 to avoid conflicting targeting restrictions during bid strategy learning.
AGENT 5 — LANDING PAGE CRO ANALYST
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/05_landing_page_cro_analyst.md.
Responsibilities: identify landing page friction points, propose A/B tests to improve conversion rate and lead quality, and ensure message match.
Deliverable structure:
1) “Funnel’o ir LP diagnostikos checklist” (15 items: speed, mobile UX, form friction, trust, clarity, analytics)
2) “Top 10 trinties taškai (tipiniai) ir kaip juos patikrinti” (table: Friction, Symptom, Evidence to collect, Fix idea)
3) “A/B testų planas” — propose 6–8 tests with templates:
- Test name, Hypothesis, Page(s), Variant A vs B description, Traffic split, Primary KPI, Secondary KPI, Guardrails, Minimum sample/runtime rule-of-thumb, Implementation notes (GTM/Optimize alternative), Rollback plan
4) “Lead quality loop” (how to connect CRM outcomes back to ads; define MQL/SQL proxy if unknown)
5) “Message match matrix” table mapping: Intent theme (from Agent 3) → Ad promise → LP section → Proof (testimonial, numbers, guarantees) → CTA
6) “Rizikos / prielaidos” (≥5) + mitigations
7) “Kas gali nepavykti?” (≥5) + detection signals
8) “Įžvalgos kitiems agentams” (8–12 bullets)
9) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must coordinate with Agent 6 on event tracking requirements before recommending variants that change form flows.
- Must coordinate with Agent 1 to avoid running multiple LP structural tests simultaneously on the same traffic source.
AGENT 6 — MEASUREMENT & READOUT ANALYST
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/06_measurement_readout_analyst.md.
Responsibilities: define measurement plans for each experiment type and produce a readout framework with statistical/decision guidance.
Deliverable structure:
1) “Tracking architektūra (target state)” (GA4 + Google Ads + GTM + CRM; bullet architecture)
2) “Konversijų žemėlapis” table with 8–12 events: Event name, Platform (GA4/Ads/CRM), Type (macro/micro), Counting method, Value, Notes
3) “Eksperimentų matavimo planas”:
- For Search/PMax bidding, creatives, audiences, LP tests: specify primary metric, attribution approach, windows, and what to hold constant
4) “Statistinė / sprendimų metodika (praktinė)”:
- Provide rules-of-thumb for sample size, minimum detectable effect, and when to call winners
- Include at least one approach for low-volume scenarios (e.g., Bayesian/credible intervals, directional decisions, sequential testing guardrails)
5) “Readout šablonas (copy-paste)”:
- 1-page template with sections: Hypothesis, Setup, Dates, Results table, Decision, Learnings, Next test
6) “Dashboard & alerts” (what charts and anomaly alerts to set up; 10 bullets)
7) “Rizikos / prielaidos” (≥5) + mitigations
8) “Kas gali nepavykti?” (≥5) + detection signals
9) “Įžvalgos kitiems agentams” (8–12 bullets)
10) “Klausimai / kas man trūksta” (5–8 bullets)
Dependencies:
- Must read Agent 1 backlog and specify measurement per top 8 experiments.
- Must flag any experiments that are not measurable given typical constraints and propose alternatives.
EXECUTION ORDER (STRICT)
1) Produce 00_context_assumptions.md first.
2) Draft agents 2–6 in parallel, but do not finalize KPI thresholds until Agent 6 completes measurement rules.
3) Agent 1 finalizes the backlog + calendar only after agents 2–6 provide overlap/conflict notes and Agent 6 provides decision methodology.
4) Produce synthesis as the final step.
FINAL SYNTHESIS / REVIEW (MUST END THE PROJECT)
Create outputs/agent_teams_demo/google_ads_pasyros_analitikas/07_synthesis_master_plan.md as the single unified plan.
Structure (use exact headings):
1) “Santrauka (200–250 žodžių)” — what will be tested, why, and expected business impact
2) “Patvirtinti faktai vs prielaidos” (two-column table)
3) “Vieningas 4 savaičių testų kalendorius” (table W1–W4; include owners, dependencies, and no-overlap constraints)
4) “Top 8 eksperimentai (vieno puslapio portfelis)” (table with crisp decision criteria)
5) “Matavimo ir ataskaitų teikimo procesas” (weekly cadence, who reviews, what decisions are made)
6) “Rizikų registras” (table: Risk, Likelihood, Impact, Mitigation, Owner)
7) “Kas daroma, jei rezultatai neaiškūs per 4 savaites” (playbook with 6 steps)
8) “Ką pateikti komandai rytoj” (operational checklist of 15 items)
Review requirements:
- Cross-check consistency: KPIs and conversion definitions must match across all files.
- Resolve conflicts between agents (bidding changes vs LP tests vs audience restrictions) and document the resolution.
- Include a short “Assumption challenges” section: list 6 challenged assumptions and how to validate them next week.
- End with “Galutinis kokybės patikrinimas” checklist (12 items) verifying tracking, naming, budgets, experiment IDs, and rollback plans.
Create your own AI agent team at Build Agents Store. Describe your business problem and get specialized agent teams with ready-to-use prompts for Claude Code.