OneDrive Duplicate File Benchmark: 102,610 Files Scanned

The pipeline first grouped candidates by byte size, then compared a OneDrive-supplied content fingerprint such as quickXorHash or SHA-1 when the provider returned one. Algorithm identity travels with the value, so unlike fingerprint types are never compared as though they were interchangeable. A filename, timestamp, thumbnail, or equal size can help a reviewer investigate but cannot promote a candidate into the exact total.

Pagination, missing hashes, changing files, incomplete searches, and provider errors must remain visible. A clean-looking result is not useful if the inventory silently omitted part of the drive. The benchmark therefore reflects completed enumeration and the conservative exact grouping that the running product exposes to its review interface.

The estimate counts only copies beyond the retained baseline in each exact group. If three byte-identical items appear together, at least one remains; the estimate considers at most two extra copies. It is therefore different from the total size of every file participating in a duplicate group.

Exact content equality still does not decide which file item is redundant. A separate path may own the active shared link, useful version history, a retention obligation, or a canonical project location. ClearCopies keeps these items visible so the user can select the keeper rather than accepting a filename-based automatic choice.

One account cannot establish a universal duplicate percentage for OneDrive users. Libraries differ by camera imports, collaboration, migrations, backups, creative workflows, file types, retention requirements, and age. The result should be treated as evidence that the method operates at six-figure-file scale, not as a promise that another account will recover the same share of storage.

It also does not measure visual near-duplicates, revised documents, recompressed media, exported Google Workspace files, or cross-provider copies. Those items normally have different bytes and require a different comparison method. Keeping them outside the exact total prevents a broad similarity claim from being mistaken for safe cleanup evidence.

Record the account scope, scan completion state, file count, total byte count, supported fingerprint algorithms, and unresolved items. Preserve aggregate output before cleanup so a later reviewer can understand how the estimate was produced. If the library changes substantially, scan again rather than applying an old decision to new files.

For a cleanup plan, add ownership, sharing, location, protected-format, and retention checks after exact detection. Revalidate every selected item immediately before a later move to recoverable trash. Detection proves content equality; review and revalidation make the operational decision safer.