One of the final phases of any system implementation is parallel testing. This is one of the most critical aspects of the process but can also be the most complex. Once implementing a new payroll system, it’s imperative for organizations to plan for parallel testing so that the implementation is as uncomplicated and problem-free as possible.
What is it?
The process of parallel testing ensures that payroll is being accurately calculated by a new system. During parallel testing, the old, or “legacy” system, is run in parallel to new payroll software and the results are compared.
How does it work?
During parallel testing, the following steps should be followed:
- All the payroll information for a selected payroll cycle (including outputs such as taxes and global benefits) should be exported from the legacy payroll system and saved
- Exported information regarding the selected payroll cycle should be entered into the new payroll system
- Run payroll using the new system
- Results from the new system can then be compared to those of the legacy system, and inconsistencies addressed
Who carries it out?
The system undergoing tests will be the new means by which payroll is calculated. So it makes the most sense that parallel testing is managed by those regularly involved in payroll responsibilities. Testing allows hands-on training and generates invaluable troubleshooting experience before the system even goes live.
Unfortunately, this can double the workload required to process payroll information into both the new and old systems. Thus, additional help may be needed during this transition phase. When needed, vendor representatives or managers within the organization that have gained more experience with the new system can help by answering questions and providing information.
Determining payroll cycles to be tested
The payroll cycles being tested should reflect normal payroll cycles. If you are running parallel testing during cycles in which there are numerous salary increases, bonus payments, or large intakes of new hires, the likelihood of errors may increase depending on your process and levels of automation.
Also keep in mind that running parallel testing during cycles in which there are less than the standard number of payroll transactions, such as the holidays, may not give an adequate representation of how well the system will perform during normal payroll cycles.
How many payroll cycles should be tested?
I’d go with one to three, but this number will be dictated by such factors as the volume of employees, levels of complexity, amount of payroll automation, etc. Organizations have different needs and capabilities, so the optimal time and methods for parallel testing may vary.
Categorizing discrepancies
Once parallel testing has been run for one payroll cycle, the results from the new and legacy systems should be compared. Every discrepancy found should be highlighted and categorized according to the cause of discrepancy.
Causes of discrepancy categories may include:
- Entry errors
- Explainable or acceptable differences (such as rounding errors)
- Errors in rule creation
- Unexplainable errors
After categorizing these inconsistencies, the discrepancies or the cause of the discrepancies should be corrected as soon as possible. This will allow further parallel payroll testing to produce more accurate results. If there are unexplainable errors, parallel testing should continue until the errors can be clarified and rectified. The system should not “go live” until you are certain that payroll will be 100 percent accurate.
