| name | metrikia-budget-alert |
| description | Autonomous budget and anomaly monitoring for ads via Metrikia. Detects aberrant spend, dropping ROAS, creative fatigue, and alerts only when action is required. Runs on a schedule via blueprint, delivers to the home channel. |
| version | 1.0.0 |
| author | Alexandre Mallet / BULDEE |
| license | Apache-2.0 |
| platforms | ["macos","linux"] |
| metadata | {"hermes":{"tags":["Marketing","Monitoring"],"blueprint":{"schedule":"0 */6 * * *"}}} |
| required_environment_variables | [{"name":"METRIKIA_API_KEY","prompt":"Metrikia API key (mk_live_...) for the Metrikia MCP server"}] |
Metrikia budget and anomaly alert
Autonomous watch: speak only when something needs action. No noise.
Prerequisite
Metrikia MCP server configured. Tools: get_anomalies, get_metrics, get_budget_advice, get_sync_status.
When to use
- Autonomously via blueprint (every 6h).
- On demand: "check my ad alerts".
Procedure
- Sync:
get_sync_status. If data is lagging by more than a few hours, report the sync lag and stop (no alert on stale data).
- Anomalies:
get_anomalies over the last 24h. Filter on real severity (ROAS dropping below a threshold, exploding spend, confirmed creative fatigue).
- Budget: if an anomaly touches allocation,
get_budget_advice for the concrete recommendation (cut, cap, redistribute).
- Notification decision:
- Nothing significant: do NOT notify (silence is a good sign).
- Action required: short alert + actionable recommendation.
Alert format (only if action is required)
Metrikia ad alert
[campaign/lever]: [problem] (Metrikia attribution)
Impact: ...
Recommended action: ...
Pitfalls
- High noise threshold: better to miss a micro-variation than to spam. One alert = one action.
- Never alert on a platform ROAS. Always Metrikia attribution.
- No alert on stale data (check the sync first).
Verification
If nothing is actionable, the run ends with no notification. Any alert emitted contains a concrete action, not just an observation.
