How it works
Enter the brands you care about and the query you want to win. The matrix shows simulated recommendation strength for each brand × assistant cell, with colour-coded tiers and an averaged composite column.
Why this matters
- The assistants disagree more than you'd expect - a matrix view makes the disagreements visible and turns them into a prioritised fix list.
- Wide variance across assistants for a single brand is a signal that the brand's entity model is weak - the fix is usually schema or Wikidata.
- Narrow variance at a high score is the signature of a default recommendation - you want every row you care about to end up in that state.
Close the gaps
For every row where the variance is wide, use the Recommendation Simulator to model a fix and the Recommendation Probability calculator to verify the expected lift.