Limits management could set, and see followed

The situation

From 2010 to 2014 I was a senior engineer at a software firm, building a loan origination system for a bank's commercial lending team. It was client work from the start: their process, their systems, their deadlines.

Two groups touched every loan. Loan officers in production brought deals in and quoted terms. Credit decided whether the bank should make the loan. Above both, management set the rules: how much the bank would lend against a property and its income, what rates could be quoted, and who could approve what.

The problem

Those rules lived in policy documents and in people's heads. A deal that broke policy was caught late, when credit noticed it, or not at all, and management had no easy way to know whether staff were working inside the limits they had set.

Meanwhile the same borrower and property information was typed again at each step, and the data the team needed sat in separate systems: the core banking system, market data, rate feeds and a document imaging system.

What I did

• One path from production to credit. Loan officers entered a deal once and submitted it to credit, which picked up the same record instead of a new spreadsheet.
• Policy as settings. Management's limits on loan size, leverage, pricing and approval authority became settings they controlled. A deal outside a limit was flagged as an exception and routed to someone with the authority to approve it.
• Sizing and credit tools. As the system grew, we added tools for officers to size a loan against those limits before quoting, and for credit to analyze the deal from the same numbers.
• Visibility for management. Management could see which deals were inside policy, which were exceptions and who had approved them, without asking around.
• Enter it once. I redesigned data capture so information entered once carried through the full loan process, and integrated REIS market data, Xignite rate feeds, the bank's document imaging system and daily loads from its core banking system.
• The team. I led the development team through the early build, and introduced code reviews and sprint planning.

The result

The system processed more than $1B in loans, cut funding times by 30% and reduced duplicate data entry by 70%. Management stopped finding out about exceptions after the fact.

It was also the first time I built this kind of system, and its core ideas shaped most of my later work: capture structured data from the start, make policy something the customer configures, and give the people who sell and the people who approve a clear hand-off.

Need something like this? See the Custom build package.