I prefer to put sensor data into a C struct and send it all at once, in binary. On the RX side, you can do a memcpy() from the RX buffer into a matching struct. That way, if you add / subtract sensors in your setup, you just need to adjust the struct definition. Using a struct also allows you to mix data types (float, uint16_t, etc) depending on what your sensor produces.
Also, if you tell the library how many bytes to send using sizeof() on your struct, then it will automatically be adjusted if you change its definition.
If your sender and receiver are different processor types, you do need to be careful of word size and Endianness. But, that’s easily accommodated once you know what each processor wants.