Process measures based on cost, cycle time, labor productivity, process yield, and the like are measures of process efficiency. Suppose that a goal for our order -fulfillment process is to reduce order-picking errors to one error per thousand order lines. Managing to that goal requires identification of order –picking errors in relation to the number of order lines picked. For order –picking errors that are inadvertent that is, when they happen, the picker is unaware of them measuring them requires a separate inspection to identify errors. In a random audit on a sample of picked orders, an inspector identifies errors and records them. As with delivery-time measurement, the team must think through all the uses it will make of these measurements. For a report of estimated error rate, the data needed are: number of errors and number of order line inspected. To improve process performance in this category, the data must help the team identify error sources and determine root cause. For that to occur, each error must be associated with time of day, shift, product type, size of package, etc., so that the data can be stratified to test various theories of root cause.