This paper discusses the use of control charts in monitoring customer satisfaction survey data over time. Using simulated multinomial data, the performance of the &Xmacr;S chart, the modified p chart (or &Xmacr;P chart) and the χ2 chart are compared, assuming both in-control and out-of-control situations. It was concluded that the first two charts are almost equivalent, but for practical reasons, it is suggested that the &Xmacr;P chart should be preferred. It is clearly shown that, in many circumstances, both the &Xmacr;P and the χ2 charts are not able to detect not only statistically significant changes in the satisfaction profile but also sizeable ones. Taking into account the poor individual performance of each of these two charts it is recommended that they should be operated together. However, when an out-of-control point occurs, the direction of the statistically significant change in the underlying satisfaction level can be very difficult to interpret. Hence, a new method based on the value concepts is proposed as a tool to help decision-makers in interpreting the out-of-control signals.
