Keyword Research Engine — Performance Marketing Agentic Workflow

schema_version: 1
version: "1.0.0"
id: keyword_research_engine
title: "Keyword Research Engine"
description: "Build a keyword universe from product brief. Outputs clustered keyword list with intent classification, match types, and Keyword Opportunity Scores."
overview: |
  Starting from a product or offer brief, this chain constructs the full keyword universe for a paid-search account: seed expansion, intent classification, clustering, match-type recommendations, and an opportunity score derived from volume, competition, and estimated cost. The output is structured so it can be pasted directly into Google Ads or a bid-management tool.

  Run it when launching a new market, entering a new product line, or re-baselining an existing account. The resulting clusters double as the input for the ad-copy generator workflow, keeping theme grouping consistent across research and creative.

category:
  level: task
  domain: marketing
  reusable: true
params:
  product_brief: string
  target_audience: string
  competitors: string
  geo: string
steps:
  - type: llm.generate
    prompt: |
      You are a paid search keyword researcher building a keyword universe.

      **Product/service:** {{product_brief}}
      **Target audience:** {{target_audience}}
      **Competitors:** {{competitors}}
      **Target geography:** {{geo}}

      Follow the 5-step keyword research methodology:

      **Step 1 — Plant Seeds**
      Generate 15-20 seed keywords across three angles:
      - Product-based (what you sell)
      - Problem-based (what customers struggle with)
      - Solution-based (what outcome they want)

      **Step 2 — Expand**
      For each seed, generate expanded keywords using:
      - Modifier layering (buy, best, how to, near me, review, price, vs)
      - Question prefixes (who/what/when/where/why/how)
      - Long-tail variations (3+ words)
      Target 100-300 total keywords.

      **Step 3 — Classify Intent**
      Score each keyword:
      - Transactional (buy/price/cost/near me): Intent Score = 3
      - Commercial Investigation (best/review/comparison): Intent Score = 2
      - Informational (how to/what is): Intent Score = 0.5 (likely exclude from PPC)

      **Step 4 — Prioritize**
      Calculate Keyword Opportunity Score:
      KOS = (Est. Volume × Intent Score) / (Competition × Est. CPC)
      Rank and select top quartile.

      **Step 5 — Cluster & Assign**
      Group into ad group themes (10-20 keywords per group).
      Assign match types: Exact for high-intent proven terms, Phrase for controlled expansion.

      **Output format:**

      ## Seed Keywords
      [Table: Seed | Angle (Product/Problem/Solution)]

      ## Keyword Universe (top 50 by KOS)
      | Keyword | Intent | Intent Score | Est. Volume | Competition | KOS | Match Type | Ad Group |

      ## Ad Group Structure
      | Ad Group Name | Keywords | Theme | Recommended Match Types |

      ## Negative Keywords
      [Universal negatives + vertical-specific negatives]

      ## Next Steps
      - Search term review cadence
      - Expansion opportunities identified
    as: keyword_list

  - type: variable.set
    name: result
    value: "{{keyword_list}}"