TOOL

Inputs: campaign brief and optimization guardrails

Define what the agents are allowed to do before simulations begin. These inputs shape creative generation, audience selection, and budget movement decisions.

Campaign brief

Goal, budget, run window, channels, approved claims, and required brand constraints.

Audience & keyword constraints

Allowed targeting pools, exclusions, keyword sets, and negative-keyword definitions.

Economic guardrails

Target CAC, target LTV, max budget shift percentage, and confidence threshold.

Current app data

Acquisition app data

Self-contained sample data is ready

Acquisition can run immediately with campaign briefs, audiences, creatives, and simulated performance data.

Campaigns1
Audiences2
Creatives2
Performance4

Active source: current acquisition sample app tables. Sign in to apply imported workspace data.

Current table rows: 1 campaigns, 2 audiences, 2 creatives, 4 performance rows

Available imported datasets: 0

Provider snapshots: 0

Workspace source: None available

Active sourceSample Data

Acquisition sample data

Pending selectionSample Data

The tool will reset to the self-contained sample dataset.

Campaigns1
Audiences2
Creatives2
Performance4

Campaign bootstrap (also available in /acquisition/create)

Create acquisition campaign

Bootstrap the paid-acquisition loop directly in-app (no CLI required).

Channels
Audience source
Go to simulations

Assumptions behind scoring and reallocation

Model assumptions (editable in this prototype)

  • Initial budget is evenly distributed across test cells.
  • Scores combine CTR, conversion rate, CAC efficiency, and ROAS quality.
  • Bottom 30% cells donate budget to top 30% each iteration.
  • Maximum single reallocation is capped by maxBudgetShiftPct.
  • Campaign state progresses from Testing to Scaling when score quality is sustained.