This paper examines the value and impact of using purchase-level information, specific ally its role in alleviating information asymmetry in credit markets. Using a proprietary lending dataset from a firm operating in the BNPL industry, I evaluate the underwriting practices of the lender and document three findings – i) lender’s internal score predicts the likelihood of a loan defaulting with18% greater accuracy than traditional credit score, ii) lender’s internal score incorporates purchase level information, and iii) lender prices a significant portion of loans using purchase-level information. I find that a model that includes purchase-level information can lend 33 percentage points more, increasing financial inclusion without having a negative effect on the default rate. Similarly, a model that includes purchase-level information can achieve 3 percentage points lower default rate, thus increasing efficiency in this market. Leveraging a randomized experiment conducted by the lender and discontinuity in lender pricing, I find that a 10percentage point increase in offered interest rate decreases applicant take-up by 14.25 percentage points with no significant impact on the loan performance.