The engine runs a configured number of ranker evaluations across neighbor competitive sets using cached baseline target content to establish a pre-edit baseline, then repeats the process with the winning artifact to calculate the net rank change.
sequenceDiagram autonumber participant Engine as Blast Radius Engine participant Cache as Property Cache participant Ranker as LLM Ranker Note over Engine,Cache: Step 1: Baseline Evaluation Engine->>Cache: Fetch neighbor SERP & baseline content Cache-->>Engine: Competitor items + unedited target Engine->>Ranker: Rank with unedited target (repeated samples) Ranker-->>Engine: Baseline rank samples & median Note over Engine,Ranker: Step 2: Post-Edit Evaluation Engine->>Ranker: Rank with winning artifact (repeated samples) Ranker-->>Engine: Post-edit rank samples & median Note over Engine: Delta = Baseline Rank - Post-Edit Rank
Referenced by
