When overwhelmed by production, “Test” or “Quality Control” must learn to think globally rather than just functionally. Historically most companies always test first-in, first-out (FIFO) but should be prepared to abandon that practice when facing a backlog. An analogy is a navigator letting the pilot know they are off course right now versus discussing history from three hundred miles ago and working their way up to the present.
Recently a colleague was concerned that his company’s test capacity was insufficient to test all their output in a timely manner. (They manufacture integrated circuits with several hundred devices on each wafer produced.) In fact, the backlog of parts to be tested was approaching six weeks since additional test cells were not ready. After he explained the multitude of reasons why the additional test cells were not ready and that production could not be slowed to match the available test capacity, I asked how they were handling the backlog. He appeared perplexed by my question and wanted to know why it wasn’t obvious that they would simply test them in FIFO order as they had always done.
After additional discussion, the reason why they thought FIFO order was best became clear. Normally, test capacity is planned to match wafer fabrication capacity as closely as possible. As a result their metrics measure how quickly they move each lot of wafers though the test area. Beyond average cycle time, they look at the distribution of cycle times and they want to avoid any lot staying in the test area over a certain limit. Therefore to make their numbers, they need to operate as efficiently as possible on a FIFO basis.
This is completely understandable from an operations theory standpoint. However, real life rarely matches theory. In this particular case, they had optimized the test function at the expense of the overall manufacturing process. This was most likely due to their failure to examine the “big picture”-i.e. the need to look across the entire operation not just at individual functional areas.
What would make more sense in this backlog situation is to test the wafers in last-in, first-out (LIFO) order since the wafer fab needs the timeliest data possible to avoid building scrap products. Given a choice, wouldn’t you rather know how you performed this week versus how you did six weeks ago? And as sometimes happens, things may get worse before they get better. So conceivably, the backlog could easily stretch to eight or ten weeks before the additional test capacity is ready making the test results even less operationally relevant. (The test data from these older lots is still important for analyzing trends, however as the data gets older it has less influence on what the fab should do for the next lot.)
Due to product aging it may be necessary to test in a FIFO order to avoid having parts that are “stale” and have to be scrapped due to not completing the processing in time. When this is the case, it is best to simultaneously test as large a percentage of the latest lots as possible immediately while working off the oldest lots in the queue. This will still provide the fab with some guidance as to their current performance and reduce the risk of having a large backlog of scrap parts.
In summary, it is important to look across multiple functional areas to make sure they are working together as a team especially when operations are not running as “designed”. And be careful to quantify and qualify what you measure since that may influence a functional area to do the wrong thing especially in the context of the “big picture”. Knowing when to use LIFO instead of FIFO may prevent a large “dog pile” (i.e. a heap of problems or scrap) that might be called Fido.
High Technology Business Development - Ira Feldman 