Judgment, shown with its provenance.

The record before the firm

Selected engagements from twelve years leading data science and analytics inside large retail and banking organizations, before Audiences AI existed.

• Lifetime value
  $55M+ · 3% churn

  The LTV model that stopped advising and started deciding

  A customer lifetime value model existed for years as a score in a report. Rewiring it to own segmentation and treatment decisions, with humans on the exceptions, moved churn by three percent and $55M+ in incremental revenue within two quarters.

• Signal architecture
  $40M+ · +1.6% conversion

  The personalization gap that had nothing to do with the model

  Conversion was flat while the models kept improving. The rebuild went into the signal layer underneath: what the systems could actually see about a customer at decision time. Conversion moved 1.6 points, worth $40M+ incremental.

• Experimentation
  $120M

  Testing aligned to the decision calendar

  An experimentation program was re-anchored on the decisions leadership actually had to make each season, across ten-plus cross-functional teams. The winning strategies it surfaced were worth $120M, and the roadmap it produced ran three years.

• AI governance
  18 decisions automated

  An AI council that shipped instead of convening

  A governance body was rebuilt around a decision portfolio: which calls the models owned, which stayed human, and what evidence moved one to the other. Eighteen AI decisions were automated under five frameworks on a three-year roadmap.