ANSWER

Scenario
A mid-size mortgage lender complained that “simple” refinance loans were still taking 22 days on average—well above its advertised 15-day target. Management suspected underwriting staffing was the bottleneck and planned to add headcount. Before approving the budget, the analytics team ran a process-mining study on six months of event logs exported from the loan-origination system (LOS), e-signature platform, and credit bureau calls (2.4 million time-stamped events).

1. Discovering the Real Process

Process-mining software reconstructed the actual flows and overlaid performance metrics:

• Median lead time: 22 days but only 7 hours of value-adding work.

• Variant explosion: 138 distinct paths; the top 3 covered just 46 % of cases, signalling uncontrolled exceptions.

• Rework loops: 38 % of loans cycled through “Document Requested → Upload → Missing Docs” at least twice, adding 3.7 days per loop.

• Idle gaps: After “Credit Pull,” 62 % of files waited >48 hours for an initial review—even for low-risk FICO ≥ 740 borrowers.

2. Root-Cause Insights

Drilling into conformance checks showed that “Missing Docs” loops were highly correlated with loans sourced by two particular call-center teams. Text-mined notes revealed those agents were using an outdated checklist. Idle gaps traced to a rules engine that routed every file—even low-risk ones—to the same underwriting queue.

3. Re-engineering Proposal

The BA presented a two-pronged redesign:

• Straight-Through Processing (STP): Auto-approve credit pulls ≥740 FICO & <80 % LTV for an “expedited” queue; introduce robotic validations for income and asset docs.

• Checklist Modernization: Deploy an in-app wizard that enforces the current documentation checklist before a file can be submitted.

Projected benefits were simulated in the process-mining tool by removing rework loops and idle events:

The model showed a 33 % capacity release—worth more than three new underwriters.

4. Business Case & Sign-off

By quantifying each redesign lever with log-level evidence, the BA framed a clear ROI: $1.2 M annual savings (labor + lock-extension fees) vs. $300 k one-time platform changes. The executive team approved the re-engineering project, cancelling the headcount request and setting a KPI to monitor the live process via continuous process mining.

Take-away

Process-mining moved the conversation from anecdotal “we’re swamped” to data-driven proof, allowing the BA to justify a targeted, automation-first re-engineering effort that halved cycle time without adding staff.