Preservation simply means keeping data to use in future projects. When choosing data to archive you should consider what data and tools would have made your research more efficient and cost effective. What data and tools should you keep for future research - even though we do not know what the future holds.
If more data was available for reuse, research in general would be more productive. Some medical data can take up to a year to access, which is longer than the timeframe of masters and undergraduate projects, and a significant portion of postgraduate and funded research projects. You cannot reduce the time it takes to collect new data, but if the data from previous studies was available, you may not need to gather new data at all.
Sharing your data also increases the impact of your research as you will be cited by everyone who reuses your data. This will raise your research profile and may affect your employability and the success of future research funding bids. You can also rest assured that your data will be safe in the long term so your future self can also reuse your data.
If everyone deposited all of their data into an archive it would become a data cemetery instead of a resource for data reuse. Therefore, you should select data that is difficult to reproduce, if it is even reproducible, and tools that took months, perhaps years, to produce, as well as any supporting data. What data and tools would have saved you time if they were available to use when you started your project.
It is important to retain quality data and tools of independent merit, so plan to archive:
Your funding body may already have guidelines as to what you are required to deposit and where you have to deposit your data. Check these requirements when you apply for funding so you are prepare to preserve when you publish your research. Research councils in particular require that you submit data that supports your publication to an archive. The SHERPA - Juliet database provides a summary of the preservation requirements from funding bodies.