Skip to content

Paid Media Workflows

AI-powered workflows for keyword analysis and paid media performance reporting.

Prerequisites

Before starting, ensure you have:

  • Claude Code or OpenCode installed (Quickstart)
  • BigQuery access (Paid Search Search Term Report, Paid Media Keywords views)
  • Client account access

Stuck?

Running into issues? Check the Troubleshooting Guide.

Available Commands

/keyword-analysis

Purpose: Analyze paid search keyword performance to identify optimization opportunities.

What you get:

  • Keyword performance tiers (CTR + CVR pairing)
  • Search term to keyword mapping
  • Negative keyword candidates
  • N-gram pattern insights
  • Actionable recommendations

Time: 20-40 minutes (vs. 1-2 hours manual)

Data sources: BigQuery (Paid Search Search Term Report, Paid Media Keywords)

When to use:

  • Regular account optimization
  • Negative keyword mining
  • Match type analysis
  • Budget reallocation decisions

/pdm-report

Purpose: Generate paid media performance reports for weekly/monthly client updates.

What you get:

  • Executive summary (1-2 paragraphs)
  • Performance highlights by platform and campaign
  • Efficiency signals (budget pacing, cost trends)
  • Period-over-period comparison
  • 3-5 actionable recommendations

Time: 15-30 minutes (vs. 1-2 hours manual)

Data sources: BigQuery (Paid Media Daily Reporting)

When to use:

  • Weekly client updates
  • Monthly performance summaries
  • Campaign health checks

N-gram Analysis Agent

The PDM plugin also includes the Saving Ben agent for deep n-gram cost analysis:

What it does:

  • Identifies sustained structural cost issues (3+ month trends)
  • Detects emerging pressure points (1-2 month trends)
  • Recognizes team wins (improvements from active work)
  • Flags seasonal/temporary patterns
  • Groups unigram patterns to find systemic issues

Output: Markdown report + CSV export with 2-4 specific follow-up questions

Skills Used

PDM workflows automatically load:

Skill Purpose
pdm-methods Reporting patterns, CTR+CVR pairing, experimentation guardrails
writing-standards Seer voice and tone
quality-standards QA gates and evidence requirements

Core principles:

  • Always pair CTR with CVR (and CPA/ROAS where applicable)
  • Segment by business model (eCommerce vs Lead Gen)
  • Respect existing automations (Shape, TrueClicks)
  • Use Ad Variations/Experiments for testing

Installation

/plugin marketplace add seerinteractive/agents-infra
/plugin install core-dependencies@seer-agent-core
/plugin install pdm@seer-agent-core

Then run:

/pdm:keyword-analysis

Troubleshooting

Common issues:

  1. "Command not found" → See Troubleshooting Guide
  2. "No data found for client" → See Troubleshooting Guide
  3. "Query timeout" → See Troubleshooting Guide