[Previous entry: "Sample Size Considerations"] [Next entry: "Sample sigma vs. process sigma"]
06/12/2008: "OEE Calculations"
In calculating the Overall Equipment Effectiveness (OEE) for a process, should the Performance ratio calculation use all the parts, or only the good or non-scrapped parts for the ideal time calculation (numerator)? If only the good parts, wouldn't this mean that the yield affect on this process is being counted twice, once in performance and once in quality? I realize it probably doesn't matter practically if you keep the calculations consistent, but I think it could cloud the analysis a bit. For example, in a situation where the quality of a process is due entirely or almost entirely to material i.e. the performance of the process may be (close to) perfect. It is also reasonable to assume most or all of the unit cycle time was used on these scrapped parts. What do you think?
Greg D.
In practice, the Actual Time used in the Performance ratio (to compare with the Ideal Time) uses the total planned output over an actual operating time (measured directly to include the total time spent on first pass processing of both acceptable and unacceptable output). As you suggest, using only the acceptable output decreases the processing rate (less output for given processing time), in effect penalizing the Performance for poor first pass Quality. This is not necessary, since the Quality ratio already includes the impact of the first failures. For example, if the Availability and Performance were perfect (equal to 1.0), and only 70% of the planned output is acceptable in the first pass (without rework), then the Quality ratio is 70%, and the resulting OEE is 70%, to reflect that 30% of the available operating time was diverted to producing unacceptable product.
Yet, how is the additional time required for any rework activities addressed? If neither the Performance nor Availability ratios are impacted, then the OEE would be the numerically equal regardless of whether the unacceptable first pass output is reworked or scrapped. Yet, clearly the process will be unavailable for new orders during any rework activity. We could reflect the rework loss much like a Setup loss in the Availability ratio. However, when we consider the impact of the rework from both internal and external customers’ perspectives, the issue is often a loss in Performance, since the process is operational as the unacceptable Quality is detected and corrected prior to delivery. Consider also that the Ideal Time (which we compare to the Actual to calculate the Performance ratio) is used for our Lean Balancing, to ensure that downstream processes are neither starved nor overwhelmed and continuous flow is achieved. When Performance is optimized, the result is necessarily that the Ideal time is achieved and the flow is balanced. When the process is redeployed to rework output, flow is disrupted, which should be reflected as a difference between Ideal and Actual cycle times in the Performance ratio. I have seen reference to calculating a separate OEE for a rework process, but I don’t prefer that approach since the Ideal Time for the rework should be zero, and it is a diversion to consider measuring and optimizing a rework process when our focus should be on eliminating its need in the first place. Furthermore, there are many processes where the Quality suffers purely from the emphasis on Performance (we rush and make mistakes), in apparent ignorance that the two are intricately related. When the rework time is included in the total process time, the Performance ratio is rightfully impacted to indicate the real effect on the Overall Equipment Effectiveness.
pak