As infrastructure develops and adoption of smart water meters increases,
new techniques are needed to validate and learn from the large datasets
they produce. Patterns of leading digits (i.e., first non-zero digits, 1
through 9) can support this task. This study examines leading digits in
hourly smart meter readings from a western U.S. water utility with over
5,000 customer connections. Benford analysis, power law analysis, and
leading-digit-frequency analysis all indicate that the readings tend
toward values that start with 1. The findings suggest that readings from
smart water meters - and, by extension, water use by individual
customers - could be expected to follow a particular nonuniform pattern
of leading digits and that deviation from the pattern may indicate data
errors or abnormal water use. Applications are suggested for validating
water use data, comparing multiple datasets, checking projections, and
assessing meter performance. Additional work is needed to further
explore the beneficial uses of leading-digit patterns and other data
signatures in water use data from diverse datasets.