8.7. Data accuracy
Once the pilot was validated, the conversion process was tested. Moving data is easy. Validating that the data are accurate and complete is extremely difficult. (One consultant working with Huck confided that, in a previous position as a user, his company’s conversion strategy did not include a reconciliation of total customer order line items. Orders were ‘‘lost’’ in conversion and only discovered through customer complaints for non-delivery.) Huck‘s conversion strategy included numerous checks for line counts of sales orders, purchase orders, work orders, and other assorted categories of dynamic data. Inventory revaluation is a huge concern when cost accounting systems are changed. In the case of an ERP implementation, this is unavoidable (unless the cost accounting process of the legacy system is somehow duplicated). To avoid any significant revaluation issues, a process was developed to export real-time data from the converted database and compare the cost variances. The process was a relatively simple export from the legacy system of both dynamic and static data such as on hand balances, standard or average cost (Huck used both depending on the item), annual usage, average inventory, and so forth. This data was loaded into a spreadsheet with the projected cost as calculated by the new ERP system. From this spreadsheet database, numerous detailed and summary analyses were conducted. Item-to-item costs were sorted by variance, and all major discrepancies were resolved. The total inventory revaluation was less than 1%. The process was also very effective in screening and testing practice runs of the conversion procedures for accuracy and completeness. Stress tests were conducted on the system as soon as a converted database was available. The project team simulated full business operations by initiating multiple processes and performing system- intense procedures while simultaneously running CPU-intense utilities. Although no system problems were discovered during these tests, a major system response issue arose at cutover. The quality management module, which had worked flawlessly in the pilot, displayed response times of several minutes. This subsystem dictated that no inventory moves or transactions could occur until quality assurance had ‘‘signed off’’ on the transaction. The slow response time totally froze operations! Because of the extreme urgency of avoiding a complete business shutdown, there was no time to properly investigate and resolve the underlying problems. To avert disaster, the entire subsystem was uninstalled item by item and order by order.