Search Ads Performance Monitor

This drill creates an always-on monitoring system for search ad campaigns. It detects performance anomalies, identifies optimization opportunities, and generates agent-executable recommendations. This is the play-specific monitoring layer that feeds into the autonomous-optimization drill at Durable level.

Prerequisites

• Active search campaigns on Google Ads and/or Microsoft Advertising with at least 14 days of data
• PostHog tracking configured with search ad events (from search-keyword-campaign-build)
• n8n instance for scheduling automated checks
• Anthropic API key for analysis and recommendation generation

Input

• Campaign IDs for Google and Microsoft Ads
• Target CPA and target ROAS thresholds
• PostHog dashboard URL for the search ads play
• Slack webhook URL or notification channel for alerts

Steps

1. Build the daily metrics collection workflow

Create an n8n workflow using n8n-scheduling that runs daily at 9am:

Google Ads data pull:

• Query the Google Ads API for the last 7 days, segmented by day:

  SELECT campaign.name, ad_group.name, metrics.impressions, metrics.clicks,
         metrics.conversions, metrics.cost_micros, metrics.click_through_rate,
         metrics.average_cpc, metrics.search_impression_share
  FROM campaign
  WHERE segments.date DURING LAST_7_DAYS
    AND campaign.advertising_channel_type = 'SEARCH'
  

• Calculate rolling 7-day averages for: CTR, CPC, CPA, conversion rate, impression share

Microsoft Ads data pull (if active):

• Query the Microsoft Advertising Reporting API for matching metrics
• Normalize to the same format as Google data for unified analysis

PostHog data pull:

• Query PostHog for landing page metrics: search_ad_form_view to search_ad_form_submit conversion rate, bounce rate by source

Store in PostHog:

• Send daily summary metrics as custom events using posthog-custom-events: search_ads_daily_summary with properties for each metric

2. Configure anomaly detection rules

Using posthog-anomaly-detection, set up alerts for:

• CPA spike: CPA exceeds 150% of 14-day rolling average for 2 consecutive days
• CTR drop: CTR falls below 70% of 14-day average (signals ad fatigue or competitor pressure)
• Conversion rate drop: Landing page conversion rate drops below 50% of baseline (signals landing page or tracking issue)
• Budget depletion: Daily spend consistently hits budget cap before 3pm (signals underfunding of profitable campaigns)
• Impression share loss: Search impression share drops below 50% (signals bid pressure from competitors)
• Quality score degradation: Average quality score drops below 6 (signals relevance issues)

3. Build the weekly search query mining workflow

Create an n8n workflow that runs weekly using n8n-scheduling:

1. Run google-ads-search-query-mining to extract the last 7 days of search terms
2. Classify each query as: converting (has conversions), promising (clicks but no conversions yet, CPC below target), wasteful (clicks, no conversions, CPC above target), or irrelevant (clearly off-topic)
3. Auto-add wasteful and irrelevant queries as negative keywords via the API
4. Flag converting queries that are not yet exact-match keywords for human review
5. Post a weekly search terms digest to Slack: new negatives added, new keyword candidates, total wasted spend recovered

4. Build the creative fatigue detector

Create an n8n workflow that checks weekly:

1. Pull ad-level performance data for the last 30 days
2. For each ad, calculate CTR trend: compare last 7 days CTR to the first 7 days after launch
3. Flag ads where CTR has declined >30% from their initial performance as "fatigued"
4. Use hypothesis-generation to draft 3 new headline variations based on what worked in the high-performing period
5. Post fatigued ad alerts and new headline suggestions to Slack for human review (or auto-implement at Durable level)

5. Generate weekly performance report

Using posthog-dashboards, build a weekly report workflow:

# Search Ads Weekly Report — Week of {date}

## Performance Summary
| Metric | This Week | Last Week | Change | Target |
|--------|-----------|-----------|--------|--------|
| Spend | ${X} | ${Y} | {%} | ${budget} |
| Clicks | {X} | {Y} | {%} | — |
| Conversions | {X} | {Y} | {%} | — |
| CPA | ${X} | ${Y} | {%} | ${target} |
| CTR | {X}% | {Y}% | {%} | >2% |
| Conv. Rate | {X}% | {Y}% | {%} | >3% |

## Anomalies Detected
{List of any triggered anomaly alerts this week}

## Actions Taken
- Negative keywords added: {count} (est. ${savings}/mo recovered)
- Ads flagged for fatigue: {count}
- Keywords paused: {count}

## Recommendations
1. {Agent-generated recommendation with data}
2. {Agent-generated recommendation with data}

Post to Slack and store in Attio as a note on the campaign record.

Output

• Daily automated metrics collection and anomaly detection
• Weekly search query mining with auto-negative keyword management
• Weekly creative fatigue detection with headline suggestions
• Weekly performance report with agent-generated recommendations
• All data centralized in PostHog for cross-play analysis

Triggers

• Daily: metrics collection and anomaly checks
• Weekly: search query mining, creative fatigue check, performance report
• On anomaly: immediate Slack alert with diagnosis and recommended action