Operations

/ads monitor

Review pacing, delivery, tracking, fatigue, policy, and performance

Claude Code
/ads monitor

Installed as a plugin? The same command is namespaced: /claude-ads:ads monitor

What it does

Compares normalized account snapshots over time to catch pacing drift, delivery problems, creative fatigue, tracking breakage, and policy issues, while separating real anomalies from learning phases, seasonality, and reporting latency. Strictly observational: it flags and explains, it never changes anything.

Use it to

  • Run a daily or weekly account check without opening every platform dashboard
  • Verify a launch behaved as expected in its first days
  • Investigate an anomaly before anyone overreacts to it
  • Track budget pacing against plan across platforms
  • Catch tracking breakage and data-quality drift early

How it works

  1. Load comparable snapshots

    Two or more normalized snapshots with matching account, timezone, currency, metric, and attribution definitions.

  2. Validate freshness

    Data freshness and finalization windows are checked before any period is compared.

  3. Separate expected from anomalous

    Learning phases, seasonality, reporting latency, and planned changes are accounted for before anything is called an anomaly.

  4. Evaluate the account

    Pacing, delivery, conversion quality, unit economics, creative fatigue, tracking health, policy status, and changed objects.

  5. Report and persist

    Observations, confidence, likely causes, investigations, and thresholds land in a versioned monitoring bundle, with failures linked to follow-up tasks.

Reads & produces

What it reads

  • Two or more normalized snapshots of the same account
  • Compatible account, timezone, currency, metric, and attribution definitions
  • Known context: planned changes, launches, and seasonality
  • Data freshness and finalization windows for the platforms involved

What it produces

  • Observations with confidence and likely causes
  • A separation of expected effects from unexplained anomalies
  • Required investigations and decision thresholds
  • A versioned monitoring bundle with failures linked to follow-up tasks

Guardrails

What /ads monitor will and will not do, by contract. These boundaries are enforced in the skill itself, not left to judgment.

  • Never mutates the account. Monitoring produces observations and thresholds, and hands action off to optimize.
  • Refuses to compare incompatible windows or snapshots with mismatched timezone, currency, metric, or attribution definitions.
  • Does not alert on percentage swings built on trivial denominators, so a jump from two conversions to three never becomes a headline.
  • Validates data freshness and finalization windows first, because comparing unfinalized data manufactures fake anomalies.
  • States plainly when the evidence cannot distinguish noise from a material change, instead of forcing a verdict.

Questions

Why did monitor stay quiet about a 50 percent change I can see in the platform?

Most likely a trivial denominator or an incomparable window. Monitor deliberately does not alert on percentage changes built on tiny bases or on periods whose data has not finalized, because those alerts are usually noise.

Can monitor pause a campaign that is misbehaving?

No. Monitoring is observational by contract. It returns observations, likely causes, and decision thresholds, and any account change goes through the optimize draft-and-approve flow.

How often should I run it?

It is built for daily or weekly checks and for post-launch verification. Each run persists a versioned bundle, so consecutive runs build a comparable history.