When you're "living with" your data, you probably are so immersed that you do not need documentation. Everything is in your head. But if you have collaborators or assistants, they may not have the same understanding of the pieces of the project that you do. And over time, that clarity about the research project will fade. Documenting your data means to describe it clearly and completely so that you and others can use it consistently throughout the project and beyond. Sharing your data will allow others to replicate and build upon your work -- good documentation is critical for this aspect of your data as well.
"Read me" files can be created at critical points in a directory of files for humans to understand naming conventions, directory structure, data structure, and tags. This will be helpful for everyone working with the data, whether in the same institution, across the country, or viewing it years later.
Codebooks explain data files. A good codebook contains everything you would need to know in order to understand a dataset. It should supply the context of the study, timeframe and investigators, variables and their definitions, values and the explanation of each value for each variable, how missing data was handled, and methodology. Anything else needed to explain the data may also be noted.
There are programs designed specifically as electronic lab notebooks (ELN) such as Labguru and Elements ELN. Some researchers have made use of more generic resources such as Evernote and GitHub to build their ELN. The benefits include:
Best Practices for Paper and Electronic Notebooks (UW-Madison)