A recovery record is extra data written inside the archive whose only job is to let the program rebuild damaged bytes. It costs space, so the sensible question is not whether to use one but how large it should be.
What the percentage means
When you set a recovery record to 5%, roughly 5% of the archive size is added as redundant data. In practice that lets the repair recover from damage covering a comparable share of the compressed stream - not 5% of individual files, but 5% of the whole.
Choosing a number
- Archives you copy to a USB drive by hand: 3-5% is usually enough.
- Archives uploaded over a flaky connection: 5-10%.
- Long-term cold storage or optical media: 10%, plus a stored checksum.
- Multi-gigabyte backups: 3% is often the sweet spot, because repair time grows with the record size.
Test the plan, not the theory
Create a small test archive with your chosen settings, open it in a hex editor and overwrite a block from the middle. Then run the repair command. If the archive comes back, your settings are realistic for the size of files you actually store.
Recovery is not a backup
A recovery record protects against corruption. It does nothing about a deleted folder, a failed drive or ransomware that encrypts your archive along with everything else. Keep at least one copy on media that is offline most of the time.