Equivalence partitioning divides input data into groups (partitions) that are expected to be processed similarly. The key principle is that testing one value from each partition is sufficient to find defects affecting that partition, rather than testing every possible value. This makes testing more efficient while maintaining good coverage.