Takes target-deal inputs (sector, size, payer mix, hold, thesis) and returns the closest corpus comparables ranked by composite similarity score, with percentile ranks and a summary benchmark block. More focused than the IC-Memo generator — purely peer identification.