Legacy System Modernization for Civil Service Payroll
The civil service payroll system was developed 15 years ago using COBOL and a hierarchical database. It processes payments for 25,000 employees monthly. The system is stable but costly to maintain due to retiring COBOL programmers. The government plans to migrate to a modern web-based system with a relational database. A pilot migration of 1,000 employee records revealed data quality issues: 5% of records had missing bank account details, 3% had duplicate entries, and 2% had inconsistent job codes. The payroll must run without interruption, and any errors in pay could cause serious problems. The project manager must decide on a migration strategy: big-bang, phased, or parallel run. The budget allows for a 12-month project with a team of 10 developers, 2 business analysts, and 1 database administrator.
1. Recommend a migration strategy and justify it, considering risks and constraints.
Read the worked answer
Given the criticality of payroll and the data quality issues, a parallel run strategy is safest but resource-intensive. A phased migration by employee groups (e.g., by ministry or location) balances risk and resource use. Big-bang is too risky due to potential data errors and system downtime. I recommend a phased approach: first, clean and validate data for a pilot group (e.g., one ministry with 2,000 employees) while running legacy in parallel for that group for two pay cycles. This allows comparison of outputs and correction of data issues. Then expand phase by phase. With 12 months, we can do 4 phases of ~6,000 employees each, each phase taking 2-3 months including parallel run. This approach reduces risk, allows learning, and maintains payroll continuity. The team size is adequate if we automate data validation and use ETL tools. Business analysts will focus on data cleansing rules. A parallel run for the entire system would double workload and is not feasible with the team, so phased parallel is optimal.
2. Outline the steps to address the data quality issues before and during migration.
Read the worked answer
Steps: (1) Data profiling: analyze all 25,000 records to quantify issues beyond the pilot (missing bank details, duplicates, inconsistent codes). (2) Define data quality rules: e.g., bank account must be 11 digits, job codes must match valid list, no duplicate employee IDs. (3) Data cleansing: for missing bank details, cross-check with HR and bank records; for duplicates, merge or remove after verification; for job codes, map to new standard codes. Involve business analysts and HR. (4) Data validation during ETL: implement automated checks that reject records failing rules, with exception reports for manual review. (5) Pilot cleansing: clean the pilot group first, migrate, and compare payroll output with legacy. (6) Iterative cleansing: as each phase progresses, cleanse that group's data. (7) Post-migration audit: after each phase, reconcile totals and individual payments. Document all changes for audit trail. This systematic approach ensures data integrity and builds confidence.