The production background to satisfaction also comes into focus when you look at what should be sampled. For statistical quality control of production machines, you take a sample of items that have been produced and test each for conformance.
For customer satisfaction, the equivalent is sampling by product purchased, or for service industries, by service event. However, particularly in business markets, this can cause sampling problems in that a single customer may buy several products or make use of a service several times. If you sample these customers regularly, you are likely to face rapidly diminishing customer survey response rates and soon have no data.
The compromise is to take a sample of customers and to ask about their experiences over a certain period of time. Unfortunately for process control, this feedback typically just reflects the average view and will miss any key extremes that are important from an operational view.
For businesses, it is also very likely that you will have a small number of large customers and a large number of small customers. As you are trying to judge quality of delivery, clearly interviews with larger customer are of more importance than smaller customers, yet typically satisfaction is biased toward the views of the many and not the few. Indeed, it is also likely that the way in which you deliver to your largest customers is different to the way in which you service smaller customers, for example, an account team, specialist logistics, and custom builds.