The Frame of Reference (FOR) compression scheme for numeric values: Instead of compressing the actual value, use a value close to all others in the same range (for integers - often the minimum value, or the minimum without outliers/exceptionals) and encode all values using their difference from this reference. The differences typically need less bits to represent.

One could think of this as an approximation of the data by a constant + residuals.

Note: You don't use the difference between consecutive values, only from the single FOR value - which might not even be present in the pre-compression data.

The Patched Frame of Reference (PFOR) approach is a specific implementation of an improvement on this scheme.