How to choose a duplicate finder app for Mac: A transparent evaluation methodology

1. Create a new macOS user or use a fresh session. We recommend running the test on the latest macOS on a physical device. Do not use a virtual machine, it distorts the results.
2. Make a note of the configuration:
   • macOS version: _______________
   • Hardware (CPU, RAM, disk type): _______________
   • Test date: _______________
3. Quit all unnecessary applications.
4. Open Activity Monitor to track RAM usage.

1. Make a note of app details:
   • App name and version: _______________
   • Source (App Store / website): _______________
   • Price / monetization model: _______________
2. Install the app
3. Launch it and grant Full Disk Access (System Settings → Privacy & Security → Full Disk Access)
4. Check support:
   • Apple Silicon support: Native / Rosetta / unknown
   • Recent app update: _______________

All test datasets will be placed in an isolated folder:

~/DuplicateTest/

~/DuplicateTest/
├── Dataset_01_AccuracyTest/    # 851 files, 230х3 duplicates, the rest are traps
├── Dataset_02_LargeFiles/      # 50 files × 1-30 MB
├── Dataset_03_Photos/          # 10,000 photos
├── Dataset_04_StressTest/      # 200,000 files
├── Dataset_05_DeepNesting/     # 1000 files, 15 levels
├── Dataset_06_EdgeCases/       # Non-standard cases
└── logs/

Dataset 1 — Accuracy Test

• 851 files: 690 dublicates (230 groups × 3) + 161 traps
• Traps: similar names, different sizes, different dates, different metadata
• Generates manifest.json for automated result validation
• Expected result: 230 dublicates groups, 3 files per group, 0 false positives

Dataset 2 — Large files

• 200 files: 50 originals × 4 copies (originals + copy_A + copy_B + copy_C)
• File size: 1-30 MB (dmg, mp4, zip, mov, pkg)
• Dataset size: ~3.1 GB
• Expected result: 50 groups по 4 files

Dataset 3 — Photos

• 10,000 files: 5000 groups of 2 files
• Folder A: 2500 photos + 2500 APFS-clones (cp -c)
• Folder B: 2500 photos + 2500 regular copies in subfolder C
• File size: ~500 KB (JPEG imitation)
• Expected result: 5000 groups of 2 files

Dataset 4 — Stress test

• 200,000 files: 1000 unique × 200 copies
• File size: 20 KB (txt, dat, bin, tmp, log)
• Dataset size: 4 GB
• Expected result: 1000 groups of 200 files

Dataset 5 — Deep nesting:

• 1,000 files: 100 unique × 10 copies
• Nesting: up to 15 levels, including hidden folders (.hidden, .cache)
• File size: ~50 KB (txt, pdf, jpg, png, docx, mp4, zip, json, xml, html)
• Expected result: 100 groups of 10 files

Dataset 6 — Edge Cases:

• 19 files: hidden (by name and flag), symlinks, hardlinks
• 2 .app bundles, files with extended attributes (xattr)
• 6 zero-byte files
• Expected result: 4 duplicate groups (16 files in groups)

Dataset 7 — Location test:

One file (100 MB) in 8 folders:

• ~/Documents/dup_location_test/

  Copy

• ~/Desktop/dup_location_test/

  Copy

• ~/Downloads/dup_location_test/

  Copy

• ~/Movies/dup_location_test/

  Copy

• ~/Music/dup_location_test/

  Copy

• ~/Pictures/dup_location_test/

  Copy

• ~/Library/dup_location_test/

  Copy

  (requires Full Disk Access)

• ~/DuplicateTest/

  Copy

Expected result: 1 group of 7 files (duplicates in the system folder must be ignored)

A ready-to-use minimal dataset for quick app validation. It’s enough to check L1 criteria and basic L2 criteria.

Dataset structure:

~/DuplicateTest/
├── originals/     ← 10 original files of different types
│   ├── photo.jpg  ← photo with EXIF
│   ├── document.pdf
│   ├── video.mp4
│   ├── music.mp3
│   ├── archive.zip
│   ├── text.txt
│   ├── .hidden1.txt       ← hidden file #1
│   ├── .hidden2.txt       ← hidden file #2
│   ├── large_file.dmg     ← file >100 MB
│   ├── emoji_🎉.txt       ← unicode in the name
│   └── similar_name_A.txt ← TRAP (unique content)
│
├── copies/← copies of all files from originals
│   ├── photo.jpg
│   ├── … (all files)
│   └── similar_name_B.txt ← TRAP: similar name, DIFFERENT content!
│
├── nested/← deep nesting test
│   └── level1/
│       └── level2/
│   └── level3/
│       └── photo.jpg      ← third copy photo.jpg
│
├── .hidden_folder_A/      ← hidden folder #1
│   ├── file1.txt
│   └── file2.txt
│
├── .hidden_folder_B/      ← hidden folder #2 (duplicate)
│   ├── file1.txt
│   └── file2.txt
│
├── folder_dup_A/  ← regular folder #1
│   ├── doc1.txt
│   ├── doc2.txt
│   └── doc3.txt
│
└── folder_dup_B/  ← regular folder #2 (duplicate folder_dup_A)
    ├── doc1.txt
    ├── doc2.txt
    └── doc3.txt

Expected result:

What should NOT be found:

• similar_name_A.txt

  Copy

  и

  similar_name_B.txt

  Copy

  — different files, NOT duplicates (trap)

This dataset checks:

• ✅ Detection accuracy (1.1)
• ✅ False positives (1.2) — similar_name trap
• ✅ Hidden files (2.10) — .hidden1.txt, .hidden2.txt
• ✅ Hidden folders — .hidden_folder_A/B
• ✅ Folder duplicates (2.3) — folder_dup_A/B
• ✅ Deep nesting — nested/level1/level2/level3
• ✅ Unicode in names — emoji_🎉.txt
• ✅ Large files — large_file.dmg

Launch the app and answer:

1. Launch the app and complete onboarding (if any).
2. Find and make a note of:
   • Where folders for scanning are selected: _______________
   • Where settings are (filters, skip list): _______________
   • Where scanning starts: _______________
   • Where results and removal are shown: _______________

Why it matters: Unique files must never be marked as duplicates. A mistake here can lead to loss of important data.

How to check:

1. Open 10 random groups in the scan results.
2. For each group, make sure that the files are truly identical:
   • Same size
   • Same content (open and compare)
3. Pay special attention to Dataset 1 (traps) and Dataset 6 (edge cases).
4. Make a note of the number of false positives found: _______

Why it matters: An accidental click must not cause data loss. There must be a confirmation step.

How to check:

1. Select 3–5 files to be deleted.
2. Click Delete/Remove.
3. Is there a confirmation dialog?
4. Does it show:
   • Number of files
   • Space size to be freed
   • Where files will be deleted (Trash / permanently)

Why it matters: You must not be able to delete all files in a group — at least one must remain. Otherwise, all of the data will be lost.

How to check:

1. Find a group of 2 duplicates.
2. Try selecting BOTH files for removal.
3. Click Delete/Remove.
4. What happens?

Why it matters: After deleting a duplicate, the remaining original must not be damaged.

How to check:

1. After deleting a duplicate, locate the remaining original.
2. Open it with the appropriate app:
   • Photo → Preview or Photos
   • Video → QuickTime
   • Document → relevant app
3. Does it open fully and correctly?
4. Repeat for 3-5 various file types.

Why it matters: The app must not suggest deleting system files from ~/Library/.

How to check:

1. Scan the

   ~/Library/

   Copy

   folder.
2. Check whether the app shows files from there in results.
3. If so, try selecting a file from

   ~/Library/Preferences/

   Copy

   for removal.

Why it matters: The app must not crash, even on large datasets.

How to check:

1. Scan Dataset 4 (200,000 files).
2. During the scan, check:
   • If there are any crashes?
   • If there are any freezes (beachballs)?
3. In Activity Monitor, check:
   • RAM usage: _______ MB
   • If memory grows uncontrollably?

Why it matters: The app must run for users who don’t grant Full Disk Access.

How to check:

1. Disable Full Disk Access for the app: System Settings → Privacy & Security → Full Disk Access → toggle off
2. Restart the app.
3. Try scanning

   ~/Documents/

   Copy

   ,

   ~/Desktop/

   Copy

   ,

   ~/Downloads/

   Copy

4. See how the app behaves.

Why it matters: The app lets you skip tiny files and focus on large duplicates that actually take up space.

How to check: Is there an option in settings to set a minimum file size?

Why it matters: A user may only need one type of search — for example, file duplicates only, without similar photos. Choice saves time.

How to check: Can you choose what to search for?

• File duplicates
• Similar photos
• Similar audio

Why it matters: Sometimes entire folders get duplicated (for example, project backups). Removing a whole folder is more efficient than file by file.

How to check: Is there a mode to find duplicate folders (not just files)?

Why it matters: It lets you skip folders you don’t want scanned (node_modules, .git, caches) and speeds the search.

How to check: Can you skip folders, files, or extensions from the search?

Why it matters: If scanning is too slow, the app isn’t practical for regular use.

How to check: Measure scan time and compare with the baseline:

Why it matters: Users need to control the process. If a scan runs too long, there must be a way to stop it.

How to check:

1. Start scanning a large folder (Dataset 4)
2. Are there Pause and Cancel buttons?
3. Do they work?

Why it matters: Without a progress indicator, it’s impossible to tell if the app is working or frozen. It’s crucial for scanning large drives and folders.

How to check: Do you see progress during the scan?

• Progress bar
• Percentage complete
• Number of processed files
• Number of duplicates found

Why it matters: Duplicates often pile up on external drives and NAS. The app should handle different storage locations.

How to check:

1. Connect a USB drive and create a small test folder with duplicates
2. Does the app show it as a source ?
3. Does scanning work?
4. (If so) Check NAS/SMB.

Why it matters: The Photos.app library is one of the largest duplicate sources for regular users. The app is supposed to handle this library.

How to check:

1. Select

   ~/Pictures/Photos Library.photoslibrary

   Copy

   as a source.
2. Run the scan.
3. Does it find duplicates inside the library?

Why it matters: Hidden files (dotfiles or flagged as hidden) can also be duplicates. Skipping them means an incomplete scan.

How to check:

1. Use Dataset 6 (Edge Cases) — hidden files are in there.
2. Or create

   .hidden_test.txt

   Copy

   and a copy elsewhere.
3. Scan — are hidden duplicates found by the app?

Why it matters: Quick Look is the standard macOS preview. Its availability lets you quickly review files before removal.

How to check: Select a file in results, press Space — does Quick Look work?

Why it matters: This info helps you decide which file to keep based on directory, size, or modification date.

How to check: For each file in results, is the following shown:

• Full path
• File size
• Date modified (or date created)

Why it matters: Algorithm transparency builds trust. Users should understand why files are considered duplicates.

How to check: Do results show why files are duplicates?

• Full path
• File size
• Date modified (or date created)

Why it matters: The whole point of searching duplicates is to free up space. Users should see how much space they’ll actually free up.

How to check: Can you see how much space will be freed after deleting selected files?

Why it matters: With thousands of duplicates, manual selection isn’t realistic. Autoselect saves time and reduces mistakes.

How to check: Is there an automatic selection feature?

Why it matters: Different users have different priorities: keep the newest file, keep files in a specific folder, etc.

How to check: Can you configure autoselect rules?

• Keep newest
• Keep oldest
• Keep by path (e.g., ~/Documents)
• Keep least nested

Why it matters: Sometimes you need to see the file in its folder context — what files are next to it and what the folder’s name is.

How to check: Can you open the file or folder in Finder directly from the results?

Why it matters: It’s the last chance to review what will be deleted. It is especially important to review selected files after using autoselect.

How to check: Can you review the full list of selected files before removal?

Why it matters: Trash is a safety net against accidental deletions. But sometimes you really do need to delete files permanently (for example, sensitive data). It’s best when you have a choice.

How to check: Is there an option to either move files to Trash or delete them permanently?

Why it matters: If a file is deleted by mistake, there must be a way to recover it. The Trash provides the standard macOS recovery mechanism for that.

How to check:

1. Delete 2-3 files through the app.
2. Open Trash — are the files there?
3. Restore the file (right click → Put Back)
4. Does the file return to its original folder and open correctly?
5. Or, is there a restore command inside the app? (even better, you can immediately see “your” deleted files)

Why it matters: Errors are inevitable (locked files, no access permissions). The app should handle them gracefully without interrupting the entire process.

How to check:

1. Lock one duplicate file: Get Info → Locked ✓
2. Try deleting this file through the app.
3. What happens?

Why it matters: A good app not only reports an error but also suggests how to fix it.

How to check: When an error occurs, does the app suggest how to fix it?

Why it matters: After an operation, the user should clearly understand what actually happened — how many files were removed, how many were not, and why.

How to check: If removal is partial (some files removed successfully, some failed) — does the app show what was removed and what wasn’t?

Why it matters: Many macOS users prefer Dark Mode. A native app should automatically adapt to system appearance settings.

How to check: System Settings → Appearance → Dark. Does the app switch to Dark Mode?

Why it matters: Visually impaired users should also be able to use the app. VoiceOver support is a sign of a well-built native app.

How to check: Enable VoiceOver (Cmd+F5) and try navigating. Are key elements read out loud?

Why it matters: Not all users speak English. Localization expands the audience and improves UX for non-English speakers.

How to check: Are languages other than English supported?

Why it matters: New users may not understand how the app works. Onboarding lowers the learning curve.

How to check: Is there a first-launch tutorial?

Why it matters: If something is unclear or goes wrong, the user should be able to find help easily.

How to check: Is there built-in help, FAQ, or quick access to support (Contact Us)?

Why it matters: A good warning explains the risk and helps the user make the right decision.

How to check: Do warnings explain:

• What will happen (action)
• Why it matters (risk)
• What to do (recommendation)

Why it matters: For visual content (photos, documents), it’s convenient to compare files next to each other to pick the best one.

How to check: Can you compare two files from a group side by side in the interface?

Why it matters: Sorting by size helps remove the largest duplicates first and free up space faster.

How to check: Can you sort results by size, date, or path? Group them?

Why it matters: With thousands of results, you need a way to quickly find a specific file or folder.

How to check: Is there a search or filter field in the results?

Why it matters: EXIF helps you understand which photo is more “original” — by capture date, camera, geolocation.

Use real camera photos with EXIF data.

How to check: Does the app display EXIF data (capture date, camera, geolocation) for these photos?

Why it matters: If autoselect picked the wrong files, you should be able to quickly reset and start over.

How to check: Is there a one-click option to clear all selections?

Why it matters: Sometimes you may want to move duplicates to a separate folder for manual review instead of deleting them right away.

How to check: Is there an option to move files to a folder instead of deleting them?

Why it matters: A symlink lets you keep a file “in place” (for compatibility), but free up space — the link itself takes up almost no space.

How to check: Can you replace a duplicate with a symbolic link?

Why it matters: If two folders duplicate each other in part, it’s more convenient to merge them than delete files one by one.

How to check: Is there a feature to merge contents of duplicate folders?

Why it matters: For documentation, auditing, or analysis in other tools, exporting results is useful.

How to check: Can you export results (CSV, PDF, txt)?

Format(s): _______________

Why it matters: Scanning can take time. If you close the app, results shouldn’t be lost.

How to check: Close the app after scanning, reopen it. Are the results saved?

Why it matters: A history of actions helps you understand what was deleted and recover files if needed.

How to check: How to check: Is there a history of removed files or an activity log?

Where to find: _______________

Why it matters: If you see a file in the results that shouldn’t be removed, it’s convenient to skip it right from there.

How to check: Can you add a file or folder to the skip list directly from the results?

Why it matters: Some users have hundreds of thousands of files. The app should handle large volumes without freezing.

How to check: On Dataset 4 (200,000 files), is the app stable and the UI responsive?

Why it matters: During long scans, it’s convenient to see interim results and start working with them before the scan finishes.

How to check: During scanning, are already found duplicates displayed before completion?

APFS clones use Copy-on-Write. In this case, deleting a clone does NOT free up space. The user needs to know that.

How to check:

1. Create a file clone: select a file → Cmd+D (creates an APFS clone).
2. Scan the folder with the original and the clone.
3. Does the app warn that deleting the clone will NOT free up space?

Many users store files in iCloud. The app should properly handle files that exist “in the cloud only”.

How to check:

1. Put a file into iCloud Drive.
2. Right click → Remove Download (file remains only in the cloud)
3. Scan iCloud Drive — does the app see the evicted file?