Fintech software case study

Loan origination software that made approvals 63% faster

A regional consumer lender kept losing customers while they waited for an answer. We swapped its spreadsheets and shared inboxes for custom loan origination software. Approvals that used to take days now take hours, and every check still happens.

A person checking a bank account on a laptop while a phone confirms money received

One workflow in place of spreadsheets and email

A regional lender's slow, disconnected loan reviews now run through one connected workflow.

63%faster approvals

What we built

  • Single queueApplications sorted by how close to decision
  • Automatic checksCredit and identity checked through APIs
  • Audit trailRecords who did what, and when
  • Role based accessPeople see only what their role needs
Discuss a similar project
  • Confidentiality agreement before any detail shared
  • Reply within one business day

The problem: slow and disconnected loan reviews

Loan origination is everything that happens between a customer applying and the money landing in their account: collecting documents, checking identity and credit, working out affordability and making the final decision. At this lender, each of those steps lived in a different place. Applications came in by web form and by email, credit checks ran in a separate supplier portal, and approvals were tracked on spreadsheets and paper checklists.

An application could sit for days between one step and the next. Meanwhile, competitors were offering same day decisions and winning the customers. Reviewers spent more of their day chasing missing files than actually assessing them. Each step worked fine on its own, but none of them talked to each other.

Nine months to launch

How we worked phase by phase

Project phases in months

Map the real processWeeks 1 to 3: we sat with underwriters, analysts and branch staff to trace every handover
Design one workflowWeeks 3 to 6: one guided workflow with clear states, from submitted to decided
Build and integrateMonths 2 to 6: working parts released every two weeks, connected to credit and identity services
Pilot with one branchMonth 7: one branch team used the new platform alongside the old process
Roll out and trainMonths 8 to 9: each reviewing team moved across and trained on its own real cases
StartMonth 3Month 6Month 9
Discovery and designBuild in two week cyclesPilot and rollout

During the pilot, one branch ran the old and new processes side by side so we could confirm that decisions matched before the wider rollout.

Under the hood

How the loan origination software works

How the lending platform fits together

PeopleWho uses it
ApplicantsWeb formReviewersReact web appManagersWeekly report view
ServicesRules and checks
Key partApplication serviceNode.js workflow rulesJob queueSlow checks run hereCredit and identityChecked through APIs
Data
PostgreSQLDecisions and audit trailDocument storageSecure cloud files
Cloud
Docker on AWSAmazon Web Services

Every application now lives in one place. The moment a customer applies, the system starts the background checks, asking the credit and identity services for results through their APIs (application programming interfaces, the standard way one system asks another for data).

Reviewers work from a single queue, sorted by how close each file is to a decision. Simple cases move on by themselves once every check has passed. Cases that need a person's judgement go straight to the right reviewer with the evidence attached. Every action is saved in an audit trail, a permanent record of who did what and when.

Security and quality checks

Role based access means each person sees only what their job needs. Data is encrypted while it's stored and while it moves between systems. Before any update goes live, it runs through automated tests in a CI/CD pipeline (an automated process that tests and releases each change). See our application security services.

Measured after launch

Results after launch

Time to a loan decision (before = 100)

Before100Before the rebuild: index 100
After37After the rebuild: index 37, which is 63% faster
Old processNew platform

Loan approvals are 63% faster than before the rebuild.

What changed for the lender

  • Decisions in hours. Most customers now get their answer the same day.
  • Fewer manual errors. Automatic checks flag gaps in an application before a reviewer even opens the file.
  • One live view. Managers can see where applications are waiting and fix the slow spots.
React Node.js PostgreSQL Amazon Web Services Docker

What we learned

Most of the time saved came from redesigning the workflow itself. Putting the old steps onto new screens would have kept the old waits. Reviewers helped shape the platform from the first sessions, so the everyday problems they knew about came up long before launch.

If your own approvals get stuck in the same way, our custom software development team can map it with you. See also the software we build for finance and banking and how we keep platforms running through software maintenance and support.

Is your approval process slowing you down?

Tell us how it works today. We're happy to sign a confidentiality agreement first.