Make Sure the Number of Reads on Trend Class Is Large Enough

You should make the Number of Reads on the trend class is large enough that individual reads do not cause statistical anomalies. For example, assume that a "normal" read in a trend class is 500 kWh and, that within the trend class, there is one "abnormal" customer using 500 percent of normal:

• If the number of reads in the class is 10, the average read is calculated to be 700 kWh. The abnormal read causes a 40 percent increase in the average.
• If the number of reads in the class is 100, the average read is calculated to be 520 kWh. The abnormal read causes a 4 percent increase in the average.
• If the number of reads in the class is 1000, the average read is calculated to be 502 kWh. The abnormal read causes a 0.4 percent increase in the average.

By sizing the number of reads in a trend class appropriately, you can reduce the impact of statistical anomalies.