What Went WrongBy David F. Carr | Posted 2002-12-16 Email Print
The confectioner's holiday sales melted when it switched software in 1999. Now, a second try is succeeding. Here are the lessons learned.#2: Unentered Data">
What Went Wrong #2: Unentered Data
How could Hershey lose track of inventory so badly that it couldn't fill orders in 1999? That's one part of the story that's never been explained well publicly.
"The project really failed on a very simple thing, which is the recognition that SAP requires a lot of discipline," says Penn State's Stenger. "SAP needs to know where all the inventory is." Specifically, the problem was that Hershey had devised informal mechanisms for dealing with the tremendous buildup of inventory to prepare for the holiday rush. "Hershey had always over the years been very good at crisis management, and they would put candy everywhere they could to store it in anticipation of this peak season. They weren't used to having to tell the computer about that," Stenger says.
This "surge storage" capacity included warehouse space rented on a temporary basis, and sometimes even spare rooms within factory buildings. The problem was that these locations hadn't been recorded as storage points in the SAP data model. Before SAP fulfills a customer order, it first checks its records of available inventory, and in this case a significant amount of inventory was not where the official records said it was.
Whether the fault for this oversight lay with logistics managers for failing to provide this data or with the information technology experts for failing to identify it as a requirement, "somewhere there was a breakdown between the technical people and the logistics people," Stenger says.
Lesson Learned: Data Is King
"We'd had a real problem with inventory accuracy, and a lot of the time we didn't have the right inventory to the right place according to our records," Hershey's Miesemer acknowledged in his speech at the warehouse management conference. Fixing those problems became one of his priorities.
Steve Sawyer, Penn State information sciences and technology professor, sees this as a typical example of the data management problems that occur in many enterprise systems implementations. "ERP systems require an overarching data model, and typically the experts within the information technology department are database administrators who are used to dealing with individual projects," he says. Often, departments fail to communicate about their data requirements except in an ad hoc way, meaning that problems are not identified until after implementation of a new system.