AAOM Handbook

manage. In one case the manager of a chemical plant decided to test this principle by ‘locking up’ the Process controls where there was an historical variability of 10%. When the operators lost the ability to change any of these parameters the variability instantly dropped to less than 2%. By definition, if a Process is statistically stable, greater than 99 measurement points in each 100 will fall randomly within the historical range, even though there will be changes from point to point. The appropriate change made in response to that 1 unusual measurement can avoid the loss of Process stability and an increase in variation. Any change made in response to one of the other 99 points will tend to produce instability and increase variation. To avoid over-reaction you have to correctly recognise which 1 point in 100 needs a response. Fortunately this is made easy by some simple statistics. Unfortunately most managers ignore the statistics and frequently over react, destabilising the Process. • Specify in detail the tasks (what they are, the sequence, the dependencies and the performance specification) that are necessary and appropriate for the Process to perform as required. • Implement the tasks accurately and consistently. • Measure that the Process is performed, and performs to, specification. • Correct all variations from the Process specification but don’t change the Process. If you manage to establish a stable Process that has output variation less than your specification limits you have demonstrated an exceptional capacity for consistency, discipline and a constancy of purpose. You have earned the right to change the Process in a way that will improve its performance capability. The simple version of the formula for a stable, low variation Process is; • Understand the expectations of the Process (the specifications). When you are planning to implement a new Process the starting point for a successful outcome is to prepare a conceptual model and flow chart that defines what outputs the Process must deliver, what inputs it will require, and what must be done with those inputs in order to deliver the required outputs. The model and flow chart for any two Processes that must produce the same type of outputs from the same type of inputs will at a summary level look the same. It is only at the most detailed level of the Process design that differences may become evident. Modelling a Process to Control Variation in Work.

Following is a generic model that could be applied to managing the Production and Service work necessary for achieving performance from a Process.