This paper presents a method for approximating the current loan-to-value (CLTV) and remaining principal structures of heterogeneous mortgage loan pools. The method uses widely available public aggregate loan data instead of loan-specific data, the availability of which is highly restricted outside lenders. The model is based on a simple matrix equation for the pool's inflows and outflows as well as on a division of the pool into multiple homogeneous cohorts. The estimated structure is compared with the true structure, as reported by Finnish banks. This comparison indicates the method is accurate. The resulting CLTV and remaining principal structures help to improve the accuracy of mortgage loan credit risk models and enable a reliable approximation of the pools' expected loss given defaults (ELGDs).