After migrating data as specified in the Work Product: Data Migration Specification, you need to validate the
correctness of the resulting data. This is a critical activity. Improperly converted data can lie dormant and cause
invalid results in the new system and, often worse, invalid results that remain undetected. Careful validation is
needed to prevent this time bomb effect. The risk is often heightened because of the volume of much of the data and
because the project team may have only indirect control of the conversion process.
As with all testing activities, you should first define the test strategy that you will use to validate the migrated
data. In addition to what is described in Task: Define Test Approach, the following considerations should be
taken into account:
In data migration, the resulting data may not always need to be completely accurate, as complete accuracy may be
uneconomic or impossible. You should define the level of accuracy that will be acceptable in your context. Here are
For accounting applications, figures must be accurate but may be only at a summary level.
For inventory applications, stock records for expensive items must be exact, but low cost items can be by weight or
volume and not necessarily by unit count.
For some applications, such as large mailing list applications, it is rarely possible to transfer all source data
in fully validated output format nor to remove all duplicates. A small percentage of inaccuracy and duplication,
however, may not be a serious problem, as long as most of the data transfers successfully.
Testing of Automated Data Migration
Special attention must be paid to automatically migrated data to ensure that there are no errors in the migration
software. Migrated data should be verified to ensure that an appropriate level of accuracy
has been achieved.
When results fall outside the acceptable accuracy range, identify the causes and initiate corrective procedures, such
Make required corrections to source data and re-run the conversion.
Identify corrections to the automated data conversion software (typically by creating a Change Request), and re-run
the conversion once the software has been fixed.
Note the data errors for manual correction on the new system.
Control procedures must be defined to ensure that all input data is completely and accurately converted. These
procedures can consist of manually checking all or a sampling of data before and after conversion or manually checking
of reports created from the data. The degree of validation required depends on the criticality of the data being