Monitoring and accurately interpreting service processes with transactions that are classified in multiple categories

Abstract

Consider a process where transactions, such as customer service transactions, are classified into categories. With just two categories, the fraction in each can be monitored with the familiar p-chart based on the binomial distribution. This paper presents a new method for monitoring the number of transactions among K categories the p-tree method, which provides an accurate and easy way to help pinpoint the categories where there has been a disturbance. In contrast to the existing practice the proposed method not only signals an out-of-control situation but also helps identify which categories are causing the problem. It is shown that a K category process can be represented by a probability tree with K − 1 binary stages and hence monitored with K − 1 independent p-charts. Simulation studies show that the p-tree method is a helpful diagnostic tool and that the sensitivity is comparable to existing multinominal-based control charts.

Keywords:

• Monitoring
• signal interpretation
• multinomial process
• service process
• multiple categories
• chi-square
• probability tree
• Marcucci control chart
• independent control charts