Data Migration Verification
Rescop was contacted to provide support in a complex validation project for an Enterprise Resource Planning (ERP) system for an established Medical Device manufacturer. This complex project included different aspects from requirement scoping to User Acceptance Testing (UAT)
This blog will describe the most complex aspect of the project, execution of data migration verification within the framework of an ERP system where a variety of both GxP critical and non-critical data is housed. The pre-existing data in the old ERP system had to be migrated to the new system. All data, including Critical GxP data, was migrated using custom migration scripts. This raised the need for data migration verification.
Verifying 200.000 entries versus risk-based approach
In total, over 200.000 data entries with multiple attributes were to be migrated. Checking all 200.000 migration entries manually would consume an unfeasible amount of time and would lead to a higher risk of verification errors. The solution was to determine a risk-based approach to decrease the number of samples. This still meant that the samples taken would have to be verified manually, meaning that the files with the exported data from the old system would be displayed next to the uploaded data in the new system. But now the total sample size was substantially smaller, while still being able to deliver a high degree of assurance that the migration had gone successfully.
Bayesian inference theorem
Although data migration is very well explained in the GAMP5 second edition, details on the implementation of the risk-based approach are missing. Multiple scientific approaches are available for this purpose among which we selected Bayesian interface theorem. The reason behind this selection was that the client already had a specific policy regarding statistical methods for validation and verification within their document management system.
This theorem provides a matrix consisting of a combination of “Reliability” and “Confidence” displayed in percentages. While the calculations used for this matrix and corresponding sample sizes are incredibly complex, the basic idea of the theorem is that the likelihood of a new event is calculated using the knowledge of prior events, combined with the assumption that all events have an equal likelihood of occurring. Using the given “Reliability” and “Confidence” matrix, the QA department determined to take 11 and 59 samples for low-impact and high-impact data objects, respectively.
GxP impact assessment
The next challenge was to determine which data was of high impact and of low impact. To do this, GxP impact of the data was assessed through the key criteria including:
- Traceability of products,
- Regulatory information,
- Closing information.
In close contact with the company’s QA manager and migration team, an impact assessment was performed to identify and verify GxP impact. Accordingly, an overview of all data objects, along with the corresponding sample sizes needed for Bayesian inference, was created.
Mapping the data objects and attributes
Before starting the migration verification process, a mapping needed to be created to determine where the source data could be found in the new system. In some cases, the same data appeared multiple times on different pages in the new system and in some cases the attributes had different names. The mapping was done by manually exploring and getting familiar with the new ERP system and documenting which data was displayed on the specific pages. It was possible to do this on a test environment, as the data migration team performed test runs there before going to the production environment.
Migration verification results
A custom traceability document was made during the migration process. Any failure was immediately communicated with the data migration team to discuss whether this was an explainable failure in the event of a deviation. Afterwards, a comprehensive report was written which summarized the process and the results.
After a successful data migration followed by an effective Go-Live of the system, the company was audited with a special interest in the new ERP system. The auditors expressed specific gratitude to all validation documentation, including the data migration verification; a crowning achievement for the Rescop consultants involved in managing this challenging data migration.