Financial services
A repeatable process for mortgage and savings reporting.
Data pipelines for a UK building society, covering liquidity reporting, redemption calculations, and analysis of its mortgage book.
Discuss a similar project
The problem
Related mortgage and savings records need to support several reporting and calculation processes.
What changed
The specified reporting and redemption calculations run through defined PySpark pipelines, with a view of the mortgage book for the risk team.
Project scope and tools
UK building society
- Mortgage and savings data
- PySpark data pipelines
- Liquidity reporting and risk analysis
Client name withheld.
Several reporting needs, the same source data
The work covered reporting and calculations based on the building society's mortgage and savings systems. Outputs included a liquidity return, mortgage redemption quotes, and a view of lending risk across the mortgage book.
These tasks use related records for different purposes. A redemption quote calculates the amount needed to repay a mortgage. Portfolio analysis lets a risk team examine lending across many accounts.
What we built
We developed data pipelines in PySpark, a tool for processing data with code. The pipelines used mortgage and savings records to generate the ALMM70 liquidity return and calculate redemption quotes.
The reporting also covered loan-to-value ratios and arrears across the mortgage book. Loan-to-value compares the outstanding loan with the property value. Arrears identify payments that are overdue.
From manual preparation to repeatable calculations
The work moved the specified reporting and redemption calculations into automated processes. It also provided a view of the mortgage book for the risk team.
The useful change is that calculations can be run again through the same defined process. When reporting logic needs to change, that change has a specific place in the code.
Let's talk about your project
Discuss a similar project
Tell Sam what you would like to change. A few sentences are enough to start.
Prefer email?
[email protected]