NBFC Loan & Collection Analytics
on Apache Superset

The client

A non-banking financial company (NBFC) whose loan accounts, disbursements, repayment schedules and collections are recorded in an Oracle database. Its name is withheld under a non-disclosure agreement.

Lending teams, branch managers and management all need the same few answers, at different levels of detail: what was disbursed, what is outstanding and how overdue it is, and how much of what fell due was collected.

The brief

Those answers sit across many tables in the lending database. A single outstanding report needs loan, schedule and receipt data joined together, overdue amounts bucketed by days past due, and totals rolled up by branch and product.

The client wanted that reporting inside its own application, refreshed on a schedule for the period being reviewed, with room to add new reports without rebuilding the pipeline each time. And reporting load had to stay off the database the lending operation runs on.

The approach

We separated reporting from the system of record. Scheduled synthesis jobs read the tables each report needs from Oracle, apply the business rules, and write the results into a PostgreSQL database that Apache Superset uses as its reporting store.

Superset sits on top of that store with shared datasets and metrics, so loan, outstanding and collection figures are defined once and every dashboard agrees. Its dashboards are embedded in the client’s application, so staff see them where they already work.

Each run is parameter based. The reporting period and as-of date are inputs to the run, so the existing dashboards refresh for the new period, the periodic reports are generated from the same data, and a new report is a new Superset chart rather than a new pipeline.