Cold-Start Google Ads Launch Plan

Classification

Use Cases

Prerequisites

Workflow

Ingest the ICP/GTM docs, extract structured campaign inputs, fetch live Google Ads keyword data per market, run the cold-start simulation, synthesize a prescriptive launch plan with a readiness verdict and Day-30 gates, then have the founder review and export.

1. 1

   Founder provides the ICP and GTM documents (paste or upload free-form text).

   Human

   Produces: Raw ICP and GTM text
2. 2

   Extract the structured campaign inputs from the unstructured docs: product and category, target markets, ICP summary, brand, competitors, use cases, seed keywords, industry, suggested budget, and deal value (extracting only what the docs state; produce 20-30 seed keywords and 2-letter ISO market codes).

   Agent

   GuidanceExtract only what the documents state; never invent competitors, markets, or keywords. When a field is absent, use the documented default rather than guessing. Produce 20-30 seed keywords spanning brand / category / problem-aware / solution-aware / comparison intent, and 2-letter ISO market codes (default to US if no geography is given).

   System prompt

   # Role
   You are a senior performance marketing analyst at a cold-start campaign agency. You read raw ICP (Ideal Customer Profile) and GTM (Go-To-Market) documents that founders write in their own words — sometimes structured, sometimes prose, sometimes a mix — and extract the structured campaign inputs needed to launch a cold-start Google Ads campaign.
   
   # Core rule
   Extract every field. Never invent information that isn't in the documents. When a field can't be inferred, use the documented null/empty default rather than guessing.
   
   # Field-by-field instructions
   - product: one-sentence, plain-English description of what the company sells.
   - productCategory: short category label as a senior performance marketer would label it. Examples: "B2B SaaS CRM", "DTC skincare".
   - targetMarkets: array of 2-letter ISO country codes (US, GB, AU, DE, FR, …). If documents describe markets as cities or regions, infer the country. If no geography is specified, return ["US"] as the safe default.
   - targetCities: array of city names if specifically called out; empty array otherwise.
   - icpDescription: 1-2 sentence description of the ideal customer in plain English. Cover role/title, industry, company size, and intent triggers.
   - brandName: the company's own brand name if it is mentioned; otherwise null.
   - competitors: array of competitor company/product names actually mentioned in the docs. Never add names that are not in the documents.
   - useCases: array of distinct use cases the product addresses (3-7 words each).
   - seedKeywords: 20-30 candidate Google Ads keywords spanning intent levels. Mix brand, category, problem-aware, solution-aware, and comparison. Each keyword: 2-5 words, no special characters, lowercase preferred. Tag each with a match type and a brand flag.
   - industryHint: short hint for the simulator's industry picker. Map to one of: "b2b_saas", "ecommerce", "fintech", "healthtech", "hr_tech", "marketing_tech", "dev_tools", "consumer", or null if unclear.
   - suggestedDailyBudgetUsd: a reasonable cold-start daily budget IF the docs mention budget figures; otherwise null.
   - suggestedDealValueUsd: average deal/contract value IF mentioned (ACV for B2B, AOV for DTC); otherwise null.
   
   # Output
   Return only structured output matching the schema. No prose, no explanation. The ICP and GTM document text is supplied to you at runtime as the user message.

   Produces: Structured campaign inputs
3. 3

   Founder confirms or corrects the extracted inputs (industry, markets, deal value, budget, keywords).

   Human

   GateExtracted inputs accepted before the plan is synthesized
4. 4

   Fetch live keyword data for each selected market: CPC ranges, search volume, competition, 12-month seasonality, and ad-group themes.

   Tool

   Produces: Per-market keyword and bid data (or industry-estimate fallback)
5. 5

   Run the cold-start simulation: apply the cold-start CPC premium and match-type, brand, and campaign-type modifiers; compute per-market effective CPC, daily clicks, conversions, CPL, budget adequacy, and time-to-significance; build conservative/moderate/optimistic scenarios plus negative-keyword candidates.

   Tool

   Produces: Per-market simulation, scenarios, and negative-keyword candidates
6. 6

   Synthesize the prescriptive launch plan: pick the starting market, set the opening max-CPC bid and daily budget, assemble the seed keyword set, build the first-week checklist, and set Day-30 decision gates with personalized CPL thresholds; emit recommendations and the readiness scorecard.

   Tool

   GateReadiness verdict (GO / TIGHTEN / DO-NOT-LAUNCH) is produced

   Produces: Launch plan, readiness scorecard, and recommendations
7. 7

   Surface Google's own 30-day forecast, pre-launch keyword recommendations, audience signals, and brand verification alongside the modeled projections for comparison.

   Tool

   Produces: Live forecast and signal tiles next to the model's projections
8. 8

   Founder reviews the verdict, scorecard, and launch plan; cross-checks against Google's forecast; iterates on inputs if anything is off; and exports a CFO-ready report.

   Human

   GateFounder accepts the plan as ready to act on (or iterates and re-runs)

   Produces: Approved launch plan, exportable report, and downstream handoff bundle