Plan around the data you need to restore
Start with the logical size of the files and systems you will protect, not the capacity printed on their disks. Match the unit to your measurement: GB and TB are decimal; GiB and TiB are binary. Use the growth horizon to size for the end of the planning period.
Choose a backup model and the number of restore points to retain. For full-plus-incremental, use a representative changed-data percentage per additional point. Both destinations use this same policy. Headroom is added separately to each destination after its repository estimate.
Enable the recovery check only when you have targets and realistic sustained transfer rates. Allow for transfer efficiency rather than treating a connection's advertised speed as observed restore throughput. Changing an input clears the old calculation; changing only the display unit reformats it.
Full backups and incremental backups need different allowances
Full-only treats every retained point as a complete copy of the projected source. Full-plus-incremental starts with one projected full and adds a changed-data allowance for each further point. That allowance is an average, not a prediction of how much will change between particular backups.
These are product-neutral capacity models. Compression, deduplication, backup-chain rules, and pruning can change what your backup software actually stores. Check its retention behavior before buying storage from this estimate alone.
Recovery targets are not measured restore times
Recovery point objective, or RPO, is the amount of data age your plan can tolerate. Here, the check compares your backup interval with that target. Recovery time objective, or RTO, is the time allowed to recover. The calculator compares it with an estimated transfer time for the projected full source, not every retained backup. These terms come from NIST's contingency planning guidance.
Meeting either target is a planning result. Transfer time does not include finding the right backup, preparing replacement systems, restoring applications, or checking that services work. A missed target points to an assumption worth changing; a met target still needs a restore test.
Three planned copies do not prove a 3-2-1 backup
The source, local target, and off-site target represent three planned positions. The calculator cannot verify their physical locations, storage media, isolation, encryption, or recoverability. CISA's backup guidance describes the arrangement; capacity arithmetic cannot establish that you have implemented it.
Check the actual target placement and access paths, then restore representative data. CISA's ransomware guidance explains why isolated backups and testing matter.
Calculations run in your browser. Explicitly sharing a result creates a URL containing the normalized inputs, so anyone receiving that link can read them.