Correspondent Behavior Mapping: Uncovering a Pricing Engine Flaw
Designed and ran a primary research program across 400-plus correspondent banks to surface a behavioral pattern distorting JPMorgan's pricing model.
Priyansha Sharma
Associate at JPMorgan Chase & Co.
From their time as
Associate
JPMorgan Chase & Co. • 2023
Overview
Priyansha took on this research at a point where the prevailing assumption inside the team was that the top ten correspondents, who made up 60 to 70 percent of the data, all behaved in essentially the same way. Her own read of the data suggested otherwise.
The Story
Priyansha took on this research at a point where the prevailing assumption inside the team was that the top ten correspondents, who made up 60 to 70 percent of the data, all behaved in essentially the same way. Her own read of the data suggested otherwise.
She designed a 20-question questionnaire structured across three priority tiers, so that field researchers could complete the most critical questions even in short stakeholder windows. She coordinated with internal research, modeling, and business teams to pressure-test the question set before a single interview was run.
She then set up and managed a three-person field team deployed across correspondent banks in the US, coordinating across time zones and collecting qualitative data from over 100 to 150 stakeholder interviews within a single month. When the data showed convergence in some segments and divergence in others, she sent the field team back with deeper follow-up questions to understand the segmentation.
Turning Qualitative Data into a Quantitative Finding
The core analytical challenge was converting qualitative interview data into a quantitative benchmark the modeling team could act on. Priyansha structured the findings to identify behavioral patterns across all 400-plus correspondents.
The finding was significant: 25 percent of correspondents were window-shopping on Chase's price, rejecting bids not because the price was unattractive but because they had no intention of transacting. Every rejection was feeding the pricing engine as a genuine price signal, systematically biasing the model.
Convincing Stakeholders Who Disagreed
The initial stakeholder reaction was skepticism. Senior leaders who had direct relationships with correspondent counterparts believed those counterparts were acting in good faith. Priyansha held her position and presented sensitivity analysis showing the model impact of the biased dataset, demonstrating that even when Chase offered its most attractive price, the window-shopping segment did not convert.
She presented the full findings to CFOs of CCB and JPMC, area product owners, and MDs, carrying the case end to end. The analysis was validated and incorporated into the pricing engine.