A lookup table can be used to compress a sparse distribution of large numbers. For example, if you only have to deal with 256 different 16-digit numbers then you write those values into a table and represent the values with the 1-byte index into the table. For a small sequence of large numbers the overhead of the table swamps the benefits of the compression. You can readily compute how many large numbers need to be compressed before you start to see benefits.
The lookup table approach can still be used even when there are more large numbers than slots in the table (see
Hash function).
Other compression techniques share that limitation: there are conditions where the compression technique actually increases the size of the message. Google is your friend:
Data compression.
@Banfa
For example, you may have heard that it is customary to add a CRC value to a data message to insure its integrity. The CRC value is much shorter than the message ... is this an example of the compression you seek? Consider a 1 KB message. That's 1024 8-bit bytes, or 8192 bits; therefore there are 2^8192 different messages that could be sent. A 16-bit CRC has 2^16 different values. This means that on average there are 2^8176 different messages that all share the same CRC value.