# Live comparison test — Temp Cleaner vs Storage Sense vs CCleaner **Date:** 2026-08-10 · **Machine:** the developer's own daily-driver workstation (Windows 11 Pro, heavily used dev machine — your numbers will be smaller on a lightly-used PC) · **Method:** everything below was measured, not quoted from marketing. Scripts and raw snapshots are kept and the test is reproducible; claims about other products appear here only because we hold the evidence for each one. A note on spirit: this is not a takedown. CCleaner and Storage Sense are tools with different design goals. This document exists because our README makes comparative claims, and claims require a protocol. --- ## Part 1 — Storage Sense coverage (the 0.2 % number) **Method:** a read-only script enumerated (a) every location Temp Cleaner finds with its default 7-day age limit, and (b) every category Windows built-in cleanup covers (registry `VolumeCaches` — 31 categories, all of them system areas, none targeting third-party application caches). The script deleted nothing. **Result:** | | Locations | Junk (7-day+) | |---|---|---| | Temp Cleaner sees | **440** | **31.3 GB** | | of which Windows built-in tools cover | **3** | **58.5 MB (= 0.2 %)** | The remaining 99.8 % — browser caches, Electron apps, package managers, sync clients — is invisible to the built-in tools. Top three finds on this machine: a package manager cache (11.4 GB), a Store-container mirror of the same cache (10.5 GB), a sync client's temp (6.6 GB). **Bonus finding:** Storage Sense was enabled on this machine, but its schedule was the default *"run only during low free disk space"* (registry value `2048 = 0`). In practice "Windows cleans automatically" often means "not until the disk is nearly full". ## Part 2 — CCleaner 6.41 live audit **Method:** official installer (`ccsetup641.exe`, 84.3 MB, Authenticode valid, Gen Digital Inc.), silent install, **registry snapshots taken before install, after install, and after uninstall**, then compared. **What we confirmed in CCleaner's favour** (our own README table had to be corrected — it was unfair to the competitor before this test): - **Analyze IS a real preview** — 10.04 GB found in 4.6 s, categories with sizes and file counts, "No files deleted yet". Our table now says "✅ Analyze". - **It is faster than us:** 4.6 s for its fixed list vs our 93 s full walk (which found 31.3 GB). A flashlight is faster than exploring the jungle; we simply cover different ground. **What the audit found on the other side:** - **No age filter** (our slider has no equivalent), **no explanation** of what was skipped or why, granularity is categories — not concrete folders you can open and inspect. - **Default checkboxes come pre-enabled** for *Empty Recycle Bin*, *Cookies* and *Memory Dumps* — one "Run Cleaner" click executes all of that at once. - **Install footprint:** two scheduled tasks created immediately (`CCleanerCrashReporting`, `CCleanerSkipUAC`), plus the usual registry presence and in-app upgrade/updater promotions. - **After uninstall:** the visible Piriform keys were removed cleanly, **but two hidden GUID-named registry keys remained** (in `HKLM\SOFTWARE\` and `HKLM\SOFTWARE\Classes\`), both containing a machine ID and a DPAPI-encrypted blob, plus two orphaned CLSID entries. They resisted normal deletion and required elevated `reg delete /f`. **Temp Cleaner's footprint, same protocol:** zero registry keys, zero scheduled tasks, zero services — delete the exe (and its `%LOCALAPPDATA%\TempCleaner` folder, which the app itself removes when you switch to portable mode) and nothing is left. ## Part 3 — where Windows wins (fairness both ways) Recycle Bin, Windows Update leftovers / WinSxS, Previous installations — the built-in tools clean these and Temp Cleaner **never will, by design**: Windows knows its own internals better than any third-party tool, and our "nothing is lost" promise does not mix with system-area deletion. Temp Cleaner is a companion to the built-in tools, not their replacement. --- *Protocol artifacts (scripts, registry snapshots, raw JSON) are archived in the project's tooling folder. If you want to reproduce Part 1 on your machine — ask the author at [claude.ai](https://claude.ai) with this repository's link; the script is read-only and safe to share.*