Statistical approaches for analyzing customer complaint data to assess aesthetic episodes in drinking water

Customer complaints and concerns about their drinking water quality provide valuable information for assessing potential risks when utilities routinely analyze these data. This research presents strategies to evaluate complaint patterns for diversity and frequency of descriptors and to apply exceedance frequencies to evaluate when the complaints were indicative of a water quality problem. Mean complaint frequency ranged between 0.2 and 2.6 complaints per 1,000 customers per year for different utilities. The most common temporal pattern was a shift to lower complaint frequencies on weekends. Weekday and seasonal patterns were slight and varied by utility. Autocorrelation analysis indicated little predictive ability in the number of complaints from one day to the next. Exceedance probabilities, however, could be used to set alarm values indicating atypical complaint frequencies. A combination of high frequency of complaints together with consistency of descriptors, that is, low diversity, was indicative of episodic water quality problems.