Skip to content
Campaign Intelligence

Ask why—and keep the answer tied to the campaigns.

Campaign Analyst works from a frozen set of campaigns, dates, metrics, and comparisons. Every finding keeps its source evidence and data limitations visible.

Campaign Analyst is read-only. It can create a proposal for review, but it cannot change the Meta account.

01

A defined question

02

A frozen snapshot

03

A traceable finding

Campaign data layers connected to one traceable evidence object.
Start with scope

A useful answer begins with a stable question.

Choose the campaigns, reporting period, comparison period, and analysis recipe. AdFlow records those inputs before calculations begin.

Input scope

Campaign selection

  • Campaign names and IDs
  • Current and comparison dates
  • Metric values and data-through time
  • Analysis recipe and operator question
Analysis receipt

Immutable snapshot

  • Scope ownership checked
  • Campaign matrix preserved
  • Tool steps recorded
  • Partial failures retained
Build the evidence

Calculations establish what happened first.

Contribution, efficiency signals, allocation outliers, and supported breakdowns are calculated before any narrative interpretation is added.

01

Scope

Selected campaigns, dates, and analysis recipe.

02

Snapshot

Fixed campaign data and data-through time.

03

Calculate

Contribution and supported performance signals.

04

Review

Facts, interpretation, and limitations.

05

Follow up

Questions against the same frozen snapshot.

06

Recommend

An editable proposal with no Meta write.

Six connected analysis stages moving from campaign scope to an editable recommendation.
Review the result

Fact, interpretation, and proposal are not the same thing.

AdFlow labels each layer so an operator can inspect what was measured, what was inferred, and what is only being suggested.

Fact

A metric observed for a named campaign and period.

Interpretation

A bounded explanation tied to the supplied evidence.

Proposal

An editable recommendation with risk and guardrail fields.

Optional AI layer

The evidence does not depend on an AI provider.

Application logic handles the deterministic analysis path. OpenAI or Gemini can add a narrative explanation when configured.

Deterministic core

Campaign facts stay available

Provider availability does not control the underlying calculations.

Configured provider

Narrative interpretation

Relevant campaign context is sent to the selected provider. Meta access tokens are excluded from the prompt.