How metrics and AI optimization work

Orientation

Metrics show what happened in your campaigns. AI optimization helps turn those signals into recommendations or, when the client has authorized it, actions that match the configured mode. Agencify brings these two layers together without hiding where the data came from or what was decided.

Why it matters

A single number rarely explains performance. The client, platform, period, objective, and data availability all change the interpretation. By keeping that context visible, Agencify helps your team compare related signals before accepting a recommendation or explaining a decision.

Mental model

Think of the flow in four parts:

• Meta or Google provides the campaign data available to the integration.
• Agencify syncs and organizes metrics, costs, results, and derived rates for the selected client and period.
• AI analysis reads the snapshot available in Agencify and produces recommendations with context and evidence.
• History records decisions and actions so your team can review what happened later.

In Alerts only mode, AI recommends and a person decides whether to apply or dismiss a supported action. In Full autonomy mode, AI can apply supported critical actions to the provider. Authorization is set for each client, and a recommendation does not guarantee a future result.

Realistic scenario

A team monitors two campaigns for the same client. One has efficient cost and consistent volume. The other has lost results in the recent period. Before acting, the operator confirms the client, platform, and date range, compares related metrics, and opens the recommendation to understand its reason and evidence.

If the client uses Alerts only, the team can apply or dismiss the recommendation. If it uses Full autonomy, a supported action may already have been carried out. In both cases, history lets the team confirm the decision, the affected campaign, and the recorded outcome.

Main capabilities

• Campaign and experiment metrics in the context of the client, platform, and period.
• Derived indicators such as rates, costs, and results when the required data is available.
• Grouped recommendations with reasons, priority, and evidence for review.
• Manual application or dismissal of supported actions in Alerts only mode.
• Automatic application of supported actions when Full autonomy is authorized.
• Latest analysis status and history for decision auditing.

Limits and recovery

Data can arrive late, vary across platforms, or become incomplete when an integration loses access. Analysis uses the snapshot available in Agencify, not a fresh direct provider read for every request. If reliable data is missing or analysis fails, AI may not generate or apply a recommendation.

When something looks inconsistent, confirm the client, platform, and period, check the integration status, and compare the latest update. Review history before repeating an action. If the difference remains after a new sync, contact support with the affected client, platform, and period.

Up next

Read **How reports and scheduling work** to turn metrics into a shareable narrative. Also see **What Agencify reads, syncs, and changes** to understand the boundaries between reading, synchronization, and a real provider change.