Google Ads Monitor

Google Ads Monitor

A recurring weekly check that makes sure your ads are still running smoothly. It catches the things that break quietly — a disapproved ad, tracking that stopped firing, a campaign that started overspending — and reports week-over-week so you see trends, not noise.

Requires: Cogny Agent subscription ($9/mo) — Sign up

Designed to be scheduled weekly via /loop or /schedule.

Works with

Call list_integrations once. Run steps 2–6 for every connected ad platform and report each in its own block. Never add spend across platforms or currencies.

Step 0: Cogny check

If no mcp__cogny__* tools exist (no list_integrations, no cogny_status), print this and stop:

Copy code
Google Ads Monitor needs your marketing data through Cogny's MCP gateway.

1. Sign up at https://cogny.com and connect Google Ads (plus Microsoft Ads, Apple Search
   Ads, ChatGPT ads, GA4, Shopify, WooCommerce or Stripe if you use them).
2. Add Cogny to .mcp.json in your project:
   {"mcpServers": {"cogny": {"type": "http", "url": "https://app.cogny.com/mcp"}}}
3. Restart Claude Code; it opens a browser for the OAuth login on first use.

Nothing to run without Cogny.

If Cogny is connected but no search ad platform (Google_Ads, Bing_Ads, Apple_Search_Ads, OpenAI_Ads) is, say "connect Google Ads (or another search ad platform) in Cogny" and stop. A platform that is listed but unhealthy, or whose first read returns 401/permission denied, gets "reconnect in Cogny" and is skipped. With Google Ads missing, run the other platforms' branches and skip the GAQL queries.

Usage

/google-ads-monitor — weekly health check of the connected account.

For a deep one-time review, run /google-ads-audit instead. This skill is the light, repeatable rhythm; the audit is the full teardown.

Steps

Google Ads money fields are micros (divide by 1,000,000). An empty GAQL result (no rows, or the tool returns no output) means zero: no matching rows or no activity in the window. Report it as 0, not as an error.

1. Read the prior snapshot

Call read_context_node on insights/google-ads/monitor/latest (and insights/microsoft-ads/monitor/latest, insights/apple-search-ads/monitor/latest, insights/openai-ads/monitor/latest for the other connected platforms). If it exists, it holds last week's metrics — every number this week is reported as a delta against it. If it does not exist, this is the first run: report absolute numbers and note that deltas start next week.

2. Account-level performance (last 7d vs prior 7d)

sqlCopy code
SELECT metrics.cost_micros, metrics.conversions, metrics.conversions_value,
  metrics.clicks, metrics.impressions
FROM customer
WHERE segments.date DURING LAST_7_DAYS

Repeat with a BETWEEN range for the prior 7 days. Compute WoW deltas for spend, conversions, CPA, ROAS, CTR.

Other platforms, same two 7-day windows:

3. Per-campaign movement

sqlCopy code
SELECT campaign.id, campaign.name, campaign.status, campaign.advertising_channel_type,
  campaign_budget.amount_micros, metrics.cost_micros, metrics.conversions,
  metrics.conversions_value
FROM campaign
WHERE segments.date DURING LAST_7_DAYS AND campaign.status = 'ENABLED'
ORDER BY metrics.cost_micros DESC

Prior week, same campaigns (dates are the 7 days before LAST_7_DAYS, i.e. 14 to 8 days ago):

sqlCopy code
SELECT campaign.id, campaign.name, metrics.cost_micros, metrics.conversions,
  metrics.conversions_value
FROM campaign
WHERE segments.date BETWEEN 'YYYY-MM-DD' AND 'YYYY-MM-DD' AND campaign.status = 'ENABLED'
ORDER BY metrics.cost_micros DESC

Join the two on campaign.id. A campaign missing from one week had no activity that week (0).

Delivery gaps (daily rows for the current week):

sqlCopy code
SELECT campaign.name, segments.date, metrics.impressions, metrics.cost_micros
FROM campaign
WHERE segments.date DURING LAST_7_DAYS AND campaign.status = 'ENABLED'

Days with zero impressions come back as missing rows, not as rows with 0. Compare each enabled campaign's dates against all 7 days of the window; every absent date is a delivery gap.

🔴 Alert if any campaign:

• Spent >40% more than the prior week
• Dropped to zero conversions after converting the week before
• Is capped at its daily budget every day (budget-limited)
• Stopped delivering entirely (zero impressions, still enabled)
• Had delivery gaps: one or more days with zero impressions while enabled (🟡 for a single day, 🔴 for two or more, or the most recent days)

Apply the same alerts to the other platforms' campaign rows. Microsoft Ads budget limits show as ImpressionLostToBudgetPercent.

4. Ad disapprovals & policy issues

sqlCopy code
SELECT ad_group_ad.ad.id, ad_group_ad.policy_summary.approval_status,
  ad_group_ad.policy_summary.review_status, campaign.name, ad_group.name
FROM ad_group_ad
WHERE ad_group_ad.status = 'ENABLED'
  AND campaign.status = 'ENABLED' AND ad_group.status = 'ENABLED'
  AND ad_group_ad.policy_summary.approval_status != 'APPROVED'

🔴 Alert on any DISAPPROVED ad. 🟡 Flag APPROVED_LIMITED (running but throttled).

Microsoft Ads: for the top-spending campaigns, tool_get_ad_groups then tool_get_ads (account_id, ad_group_id) and flag ads whose editorial status is not approved.

5. Conversion-tracking heartbeat

conversion_action rejects metrics.conversions (PROHIBITED_METRIC_IN_SELECT_OR_WHERE_CLAUSE). Get per-action volume from customer, segmented by action:

sqlCopy code
SELECT segments.conversion_action_name, metrics.conversions, metrics.all_conversions
FROM customer
WHERE segments.date DURING LAST_7_DAYS

Repeat with the prior week's BETWEEN range. An action absent from the result had 0 conversions; an empty result means 0 for every action. To see which actions are primary (and get a row for each, zeros included), use all_conversions on conversion_action:

sqlCopy code
SELECT conversion_action.name, conversion_action.status,
  conversion_action.primary_for_goal, metrics.all_conversions
FROM conversion_action
WHERE segments.date DURING LAST_7_DAYS

🔴 Alert if a primary conversion action recorded conversions last week but zero this week — tracking has almost certainly broken. This is the single most important check in the skill.

🔴 Also alert if every primary conversion action is at 0 in both weeks while GA4 key events, store orders or Stripe payments are above 0 for the same weeks. That isn't a new break, so the WoW check above misses it, but Smart Bidding has no signal and every CPA is meaningless. Run the confirmation reads below even when both Ads weeks are 0.

Confirm it before alerting. If GA4 is connected, find the property with tool_list_account_summaries, then run tool_run_report (property_id, date_ranges for both weeks, dimensions: ["eventName"], metrics: ["eventCount", "keyEvents"]). Keep eventCount: GA4 drops rows where every requested metric is 0, so an event whose key events fell to zero would otherwise vanish instead of showing as 0. If a store is connected, compare orders: Shopify tool_list_orders (query_filter: "created_at:>=YYYY-MM-DD") or WooCommerce get_sales_report (date_min, date_max). For SaaS, count Stripe payments with mcp__cogny__mcp__Stripe__stripe_api_read (GetInvoices or GetSubscriptions, filtered to each week's created range). Ads conversions at zero while GA4 events or store orders held steady = the Ads tag broke. Both at zero = the site or checkout broke, which is worse. Microsoft Ads: a campaign whose Conversions went from >0 to 0 with steady clicks gets the same alert (UET tag).

6. Budget & bidding drift

sqlCopy code
SELECT campaign.name, campaign.bidding_strategy_type,
  campaign.bidding_strategy_system_status,
  campaign.target_cpa.target_cpa_micros, campaign.target_roas.target_roas,
  metrics.cost_micros, metrics.conversions, metrics.conversions_value
FROM campaign
WHERE segments.date DURING LAST_7_DAYS AND campaign.status = 'ENABLED'

Flag campaigns whose bidding_strategy_system_status is a LEARNING_* value (compare with last week's snapshot to spot a new learning phase), LIMITED_* or MISCONFIGURED_*, and any campaign where the target CPA/ROAS is set (non-zero) but actual is >50% off target. Portfolio strategies keep their targets on bidding_strategy, not on the campaign, so a 0 target here can mean "set at portfolio level".

7. Output

Copy code
Google Ads Monitor — [Account Name]
Week of [YYYY-MM-DD] – [YYYY-MM-DD]

⚠️  Needs attention this week
- [Critical/important items first, or "Nothing — account is running clean."]

Performance (7d vs prior 7d)
- Spend ........ $X,XXX  (+/-X% WoW)
- Conversions .. XXX     (+/-X% WoW)
- CPA .......... $XX     (+/-X% WoW)
- ROAS ......... X.Xx    (+/-X% WoW)

Campaign movers
| Campaign | Type | Spend (WoW) | Conv (WoW) | Note |
|----------|------|-------------|------------|------|

Disapproved / limited ads: [N]
Conversion tracking: [OK / BROKEN — details]

Recommended actions
1. ...
2. ...

8. Save this week's snapshot

Call write_context_node on insights/google-ads/monitor/latest (and the matching insights/<platform>/monitor/latest node for each other platform) with this week's metrics (spend, conversions, CPA, ROAS, per-campaign figures, disapproval count) so next week can compute deltas. Also write a dated copy to insights/google-ads/monitor/<YYYY-MM-DD> for history.

9. Record findings — selectively

Only file a create_finding for genuine problems (disapprovals, tracking breaks, spend spikes, budget-limited profitable campaigns) — not routine numbers. A clean week should produce zero findings.

jsonCopy code
{
  "title": "Conversion tracking stopped — 'Lead Form Submit' at 0 this week",
  "body": "'Lead Form Submit' recorded 41 conversions last week, 0 this week, while clicks held steady. The tag or GTM trigger has likely broken. Check the GA4/Ads conversion tag.",
  "action_type": "conversion_tracking",
  "expected_outcome": "Tracking restored; Smart Bidding gets accurate signal again",
  "estimated_impact_usd": 0,
  "priority": "critical"
}

Cloud uses create_ticket with the same content, fields mapped per section 2 of the routing guide.

Action types: conversion_tracking, budget_adjustment, campaign_optimization, policy_fix, bidding_strategy.

Recording findings

Use create_ticket when the workspace has it (Cloud) and create_finding otherwise (Solo). The fields differ; see section 2 of ../marketing-engineer/references/mcp-routing.md.

Critical rules

1. Trends over snapshots. Always report WoW deltas. A single week's number is noise.
2. Lead with what's broken. If nothing is wrong, say so in one line and keep the report short.
3. Tracking break = critical, always. It corrupts every downstream number.
4. Read-only. Never pause campaigns or change budgets — surface, don't act. Pause tools (tool_update_campaign_status, Bing tool_update_campaign) are used only on the user's explicit request, and the skill asks first.
5. One block per platform, each in its own currency. No blended totals.