Simply put, metadata are data about data: they are the data that characterise or identify a file. As soon as you are going to publish or archive your data, you will be asked to provide these data.
Metadata serve two purposes:
Metadata are usually recorded during the archiving and/or publishing of the data. If you wish to deposit a dataset in a repository, you will be asked to provide information on the dataset. For example:
There are various metadata standards. The best-known and most used is Dublin Core. UvA/AUAS figshare also uses this metadata standard.
Some equipment and software also assign metadata to a digital file automatically, so-called embedded data. For example, the data a digital camera puts into a digital picture or the information which is registered under file properties in Microsoft Office. In Microsoft Office and in software for qualitative data analysis you can also assign your own metadata to a file.
Assigning metadata means describing a dataset in such a way that it can be read by computers, e.g. to facilitate the search function on the website of a data repository. Besides this, the dataset is accompanied by documentation: information meant to be read by humans.
To clarify the distinction by means of an example: