State Segmentation Research: Optimizing Pricing Across 50 US Markets

Priyansha ran this research concurrently with the correspondent behavior mapping, both workstreams operating under tight, sequential deadlines that fed into the same model enhancement cycle.

The objective was to identify the top three performing loan segments in every US state, across eight to nine primary segment types the team had identified over three years of research. The segments included new production, TBA, and FICO-LTV grids, among others.

The analytical challenge was that segment performance could not be read from data alone. A high-income state like New York would show strong performance in a premium segment not because the segment was inherently strong, but because the demographic profile of that state made it so. Priyansha sat with the data curation team and the fair lending team to layer in geographic and demographic context before drawing any conclusions.

She also deployed a separate set of on-ground partners to validate the internal stakeholder views on specific states. Where internal teams asserted that a particular state was a high-value segment, she sent field researchers to confirm that the ground reality matched the assumption.

The output was a state-by-state segment map that fed directly into the pricing engine, allowing the team to optimize bid prices at the MSA level rather than applying a single national view.