Companies who operate call centres such as telesales, market research or utility companies all monitor the calls that their staff make to ensure that they are complying with the relevant protocols. For example, telesales companies monitor calls in order to ensure that their staff are not mis-selling products, whilst utility companies monitor calls in order to ensure that their advisors are providing good customer service. Monitoring calls is therefore an important part of a quality control procedure, but it can be costly in terms of money, time and effort.
Statistical sampling techniques can be used to calculate the minimum number of calls that a company needs to monitor, whilst still ensuring accurate quality control results. With a small amount of information about the monitoring procedure (such as the size of the population of interest, the required confidence level and the time frame of interest), a sample size can be calculated. This provides the company with the minimum number of calls required in order to accurately monitor call quality.
When selecting which calls to monitor, the company must ensure that they choose a sample that is representative of their customer population as a whole and does not introduce any biases into the results. For example, choosing calls that only occurred during normal office hours could introduce biases since it is likely that customers who phone during normal office hours are different from those that do not. They may have different types of questions for customer services or require a different level or type of service, but this information would not normally be recorded alongside the call itself. These biases can be avoided by ensuring the company uses a well-designed sampling scheme that identifies and eliminates any biasing factors to make sure the results are accurate and reliable.
Once the sample of calls have been selected and monitored, they can be used to estimate the proportion of total calls that pass the quality control procedure. Not only does the sample provide an overall estimate of quality, but it can also be used to identify if there are any particular patterns to calls that have not met the necessary standards. For example, subsequent analysis may highlight that the quality of calls reduces near to the end of shifts which might be rectified by shortening the length of shifts or introducing more breaks. Thus, the data need not just be used to simply report on past performance but to help understand the reasons behind any past poor performance and to help develop an appropriate strategy for driving improvements.
Monitoring call quality in call centres to ensure that the calls made are of sufficient quality is a necessary, but potentially expensive task. By assessing the quality of a sample of the total calls made, a company can reduce the cost of monitoring without sacrificing the accuracy of their results. Statistical models can help understand the causes of poor performance and to help improve future call quality in the most cost-effective manner.