In a data processing scenario, you need to normalize a highly denormalized dataset to improve query performance and reduce data redundancy. What normalization techniques would you apply, and how would you ensure that the normalized data remains consistent and accurate for downstream analytics?