A competitive differentiator for Agilent is our technology center where we fabricate chip wafers. These are extremely long lead-time parts (as much as 6 months), and this particular manufacturing area typically experiences the pangs of demand volatility worse than anyone else.
In early 2008, orders were looking good, and the economy hadn't "tanked" yet. If we could simply execute to the plan, we would have a successful year; critically (with new blockbuster products) and financially (with revenue and solid ROIC). Unfortunately, the year kicked off with a serious challenge when our technology center lost the formula for one of our most important and widely used chips.
Obviously, redeveloping the formula for a wafer was a job for the brilliant engineers that work at Agilent. There were, however, other questions that needed to be answered by supply chain analysts if we were to properly prepare for the worst case scenario of not recovering the formula and exhausting all of our supply:
- How much supply do we currently have?
- How long will the supply last?
- Which products are these chips used in?
- What is the revenue associated with those products?
- Who are the customers who ordered those products? Which customers are our highest priority?
This is where the chaos ensued. Agilent has a broken demand plan. We've outsourced large portions of our supply chain to massive contract manufacturers such as Venture and Jabil. Normally, chasing a where-used in the Oracle demand plan using demand pegging should be as simple of a task as anyone can ask for. Unfortunately, due to the CM supply chains being completely unavailable, we don't have demand pegging turned on for many of the parts in Oracle, and most demand pegs to a forecast we receive from the CMs (named the Buy/Sell Forecast).
Once management realized the folly of our ways (though they'd been warned about the risk of a broken demand plan on numerous occasions), they used the "shotgun approach" to answering their questions. Quickly, large numbers of people were all working toward the same answers using slightly different approaches. If my memory serves me well, it took us approximately 2 weeks to gather answers we should have had in approximately 2 hours. The final product was the result of "interesting" data process work-arounds like using engineering BOMs (GDO in Oracle, which is complete from top to bottom) with incorrect plan factors to chase the product where-used, and then taking the products and looking up all of the orders for them in our Oracle factory ORGs. The task was tedious, the results were not extremely accurate, but it was deemed "good enough" to understand the risk of exhausting supply. In the end they redeveloped the formula, and can now build the wafer in a shorter amount of time than before the crisis.
At the time, our implementation of Rapid Response was fresh, and we didn't have the mastery of it necessary to run this type of analysis. If this happened again today, Rapid Response would be the first tool I would use because we have our major factory ORGs and our major CMs fully linked in the environment. Having complete visibility of the supply chain has made demand pegging easy again.